Abstract
Decreased erythropoietin (EPO) production, shortened erythrocyte survival, and other factors reducing the response to EPO contribute to anemia in patients who have a variety of underlying pathologies such as chronic kidney disease. Treatment with recombinant human EPO (rHuEPO) at supraphysiologic concentrations has proven to be efficacious. However, it does not ameliorate the condition in all patients, and it presents its own risks, including cardiovascular complications. The transcription factors hypoxia-inducible factor (HIF) 1α and HIF2α control the physiologic response to hypoxia and invoke a program of increased erythropoiesis. Levels of HIFα are modulated by oxygen tension via the action of a family of HIF-prolyl hydroxylases (PHDs), which tag HIFα for proteasomal degradation. Inhibition of these PHDs simulates conditions of mild hypoxia, leading to a potentially more physiologic erythropoietic response and presenting a potential alternative to high doses of rHuEPO. Here we describe the discovery and characterization of GSK1278863 [2-(1,3-dicyclohexyl-6-hydroxy-2,4-dioxo-1,2,3,4-tetrahydropyrimidine-5-carboxamido) acetic acid], a pyrimidinetrione-glycinamide low nanomolar inhibitor of PHDs 1–3 that stabilizes HIFα in cell lines, resulting in the production of increased levels of EPO. In normal mice, a single dose of GSK1278863 induced significant increases in circulating plasma EPO but only minimal increases in plasma vascular endothelial growth factor (VEGF-A) concentrations. GSK1278863 significantly increased reticulocytes and red cell mass parameters in preclinical species after once-daily oral administration and has demonstrated an acceptable nonclinical toxicity profile, supporting continued clinical development. GSK1278863 is currently in phase 3 clinical trials for treatment of anemia in patients with chronic kidney disease.
Introduction
Since its discovery as the hypoxia-induced complex responsible for erythropoietin (EPO) transcriptional regulation (Pugh et al., 1991; Semenza et al., 1991; Semenza and Wang, 1992), hypoxia-inducible factor (HIF) has been shown to recognize hypoxia-response elements in many diverse genes (Schodel et al., 2011; Semenza, 2012). The HIF family includes three hypoxia-responsive proteins: HIF1α (Wang et al., 1995), HIF2α (Tian et al., 1997), and HIF3α (Makino et al., 2001). HIFα proteins form a heterodimer with HIFβ (Hoffman et al., 1991), resulting in a transcriptionally active DNA binding complex. There is a large diversity of genes regulated by HIFα, (Schodel et al., 2011) with some cell-type and environment specificity in gene regulation (Ratcliffe, 2007; Loboda et al., 2010, 2012; Mathieu et al., 2014). Dysregulation of HIF is associated with several pathologies, including cardiac failure (Wei et al., 2012; Bishop and Ratcliffe, 2015), stroke (Philipp et al., 2006; Bao et al., 2010), lung disease (Whyte and Walmsley, 2014), and retinal damage (Semenza et al., 2000; Arjamaa and Nikinmaa, 2006).
Under normoxia, HIFα is continually ubiquitinated and degraded by the proteasome (Fig. 1). E3 ubiquitin ligase-mediated degradation of HIFα depends upon interaction with von Hippel-Lindau tumor suppressor protein (Maxwell et al., 1999; Cockman et al., 2000). The oxygen-sensing mechanism controlling HIFα stabilization involves a family of HIF-prolyl hydroxylases (PHDs) (Epstein et al., 2001), which regulate the hydroxylation of conserved proline residues in HIFα, furnishing the essential recognition element for the HIFα -VHL interaction (Ivan et al., 2001; Jaakkola et al., 2001). The PHDs are members of the iron and α-ketoglutarate (α-KG)-dependent dioxygenase superfamily and are found in overlapping but distinct tissue expression patterns (Lieb et al., 2002) with differences in subcellular localization.
The functional differences between the PHD isoforms are not well understood, but mouse genetic knockout studies suggest they have distinct functions and phenotypes (Minamishima et al., 2008; Takeda et al., 2008). In humans, a loss of function mutation in PHD2 was associated with familial erythrocytosis (Percy et al., 2006). PHD2 was found to be the primary regulator of HIF-1α levels in normoxia (Berra et al., 2003), while PHD3 was found to be the primary regulator of HIF2α, especially where PHD3 expression exceeded that of PHD2 under conditions of hypoxia (Appelhoff et al., 2004). Because the relative amounts of PHDs in EPO-producing cells in vivo are unknown, particularly in anemia, it is not clear which PHD or combination of PHDs should be inhibited to produce the best EPO-producing physiologic response. To date, no isoform-selective, 2-KG-competitive PHD inhibitors have been reported.
HIF activity is further regulated by factor-inhibiting HIF (FIH) (Mahon et al., 2001; Hewitson et al., 2002; Lando et al., 2002). FIH hydroxylates HIFα on an asparagine in the C-terminal transactivation domain (C-TAD), inhibiting the binding of HIFα to its transcriptional coactivator, p300. FIH activity is prevalent under oxygen conditions lower than those where PHDs are active; investigation of the interplay of these systems in regulation of the hypoxic response is ongoing (Chan et al., 2016).
Inhibition of PHDs results in HIF stabilization and modulation of HIF-controlled gene products, including EPO. Therefore, approaches that target HIF stabilization through PHD inhibition and the production of physiologic levels of endogenous EPO could alleviate anemia through increased reticulocyte production and subsequently increase erythrocyte levels, leading to increased hemoglobin and hematocrit levels and improved oxygen delivery (Forristal and Levesque, 2014). Furthermore, HIF affects iron metabolism modulators, including hepcidin (Yoon et al., 2006). A hepcidin decrease through PHD inhibition and hematopoiesis (Liu et al., 2012) may increase bioavailability of oral iron through increased absorption from the gut and allow better mobilization of iron stores from the liver and macrophages, potentially reducing or eliminating the need for iron supplementation often required for patients receiving recombinant human erythropoietin (rHuEPO).
Herein, we describe the discovery and preclinical development of GSK1278863 [2-(1,3-dicyclohexyl-6-hydroxy-2,4-dioxo-1,2,3,4-tetrahydropyrimidine-5-carboxamido) acetic acid], a potent PHD inhibitor that mimics the binding of N-oxalylglycine, involving chelation of the catalytic iron and blocking substrate entry. We describe how enzymatic inhibition of PHD2 and PHD3 results in stabilization of cellular HIF1α and HIF2α, leading to production of EPO and subsequent induction of erythropoiesis in vivo in preclinical species. We also demonstrate selectivity of GSK1278863 against the related enzymes collagen prolyl hydroxylase (CP4H) and FIH, and we describe an association established in nonclinical studies of structurally similar compounds between off-target activity on CP4H and development of cardiac valve lesions.
Materials and Methods
GSK1278863 was synthesized as previously described elsewhere (Duffy et al., 2007). All studies used the free acid form of the compound, GSK1278863A.
All studies were conducted in accordance with the GlaxoSmithKline Policy on the Care, Welfare and Treatment of Laboratory Animals and were reviewed by the Institutional Animal Care and Use Committee either at GlaxoSmithKline or by the ethical review process at the institution where the work was performed.
Western Blots for HIF Stabilization
The human hepatocellular carcinoma Hep3B cell line obtained from American Type Culture Collection (ATCC, Manassas, VA) was maintained in log phase growth by routine subcultivation in high glucose/glutamine Dulbecco’s modified Eagle medium containing 10% fetal bovine serum under standard culture conditions (37°C, 5% CO2).
We plated 2 × 106 Hep3B cells into P-100 cell culture dishes. After 48 hours, the dishes were treated with 100 µM deferoxamine (DFX, positive control), or 25 or 50 µM GSK1278863 in fresh medium. Dishes treated with dimethylsulfoxide (DMSO) (0.1%) were used as negative controls. The dishes were incubated for 6 hours in a CO2 incubator at 37°C. Nuclear extract was prepared using the Nuclear Extract Kit (Active Motif, Carlsbad, CA) according to the manufacturer’s instructions. Protein concentrations were determined using Bio-Rad protein assay reagent (Bradford method; Bio-Rad Laboratories, Hercules, CA) with bovine serum albumin (BSA) at standard concentrations.
Protein samples were denatured at 99°C for 10 minutes in Laemmli sample buffer, resolved on a 7% Tris-acetate gel, and transferred to a nitrocellulose membrane. The blot was stained with Ponceau S to ensure equal loading. The membrane was washed and blocked with 10% Carnation milk (Nestlé USA, Rosslyn, VA) in Tris-buffered saline containing 0.1% Tween 20 (TBST) at room temperature for 20 minutes. The membrane was then incubated overnight at 4°C on a rocking platform with mouse monoclonal HIF-1α (NB100-105; 1:350; Novus Biologicals, Oakville, ON, Canada) or rabbit polyclonal HIF-2α (NB100-122: 1:1000; Novus Biologicals) antibodies diluted in 5% milk TBST. Blots were washed in TBST 5 times and incubated for 1 hour with donkey anti-rabbit IgG (NA934; 1:2000; GE Healthcare, Piscataway, NJ) or sheep anti-mouse IgG (NA931; 1:2000; GE Healthcare) conjugated to horseradish peroxidase. Antibody-protein complexes were then detected using ECL-Plus (GE Healthcare) according to the manufacturer’s instructions.
Enzyme Assays
Materials.
LANCE Eu-W8044 streptavidin and α-[1-14C]-ketoglutaric acid were both purchased from PerkinElmer (Waltham, MA). HIF1α CAD peptide (D788-L822) was purchased from 21st Century Biochemicals (Marlborough, MA). Dansyl-GFPG-Oet and (P-P-G)10 peptides were purchased from California Peptide Research (Napa, CA). All other chemicals and materials were of standard laboratory grade.
The following recombinant proteins were used for the biochemical assays: his-strep-tev-PHD2 (1–426), his-strep-tev-PHD1 (1–407), his-MBP-att-PHD3 (1–239), his-MBP-tev-FIH (1–349), Cy5-his-gb1-tev-HIF2α C-terminal oxygen-dependent degradation domain (CODDD) (467–572), Cy5-his-gb1-tev-HIF1α CODDD (498–603), and biotin-VBC complex (composed of biotin-VHL, elongin b, and elongin c). Detailed protein expression and purification information can be found in the report by Pappalardi et al. (2008, 2011). Expression and purification of CP4H were based on the methods described by Kersteen et al. (2004) except a baculovirus infected insect cell (BIIC) expression system was used. The CP4H protein used for the biochemical assays was a tetramer comprised of full-length human collagen prolyl hydroxylase subunits A (P4HA) and B (P4HB).
LANCE Assay.
Human PHD3 in vitro enzymatic activity was measured using a LANCE (homogeneous time-resolved fluorescence quenching assay based on lanthanide chelates) format. When hydroxylated, Cy5-HIF-2α CODDD was recognized by the biotin-labeled von Hippel-Lindau protein-Elongin B-Elongin C (VBC) complex. The addition of europium (Eu)–streptavidin allowed detection of product formation by fluorescence resonance energy transfer, with a ratio of Cy5 to Eu fluorescence emission (LANCE ratio) as the ultimate readout. Assays were performed at room temperature using a 10 μl reaction volume in black 384-well Nunc nonbinding assay plates prestamped with 100 nl of compound in 100% DMSO.
The reactions were performed by dispensing enzyme (5 μl of PHD3; prepared as a 2× concentrate containing 4.0 nM PHD3, 0.2 mg/ml BSA, 400 μM ascorbic acid, 1.25 μg/ml FeCl2, 0.3 mM 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate [CHAPS], and 50 mM KCl in 50 mM HEPES at pH 7.5) into the assay plates, incubating the plates for 30 minutes, and then adding the substrate (5 μl HIF2α-CODDD; 2× concentrate contained 50 nM Cy5-labeled HIF2α-CODDD, 25 nM biotin-VBC, 1 μg/ml streptavidin-Eu, 20 μM α-ketoglutarate, 0.3 mM CHAPS, and 50 mM KCl in 50 mM HEPES at pH 7.5).
The reactions were incubated for 30 minutes and then read on a PerkinElmer Viewlux imager at an excitation wavelength of 360 nM and emission wavelengths of 610 and 660 nM. The raw data values are the quotient of the signal intensities at 660 and 610 (LANCE ratio), and inhibition in each assay well was calculated as a proportion of the high and low control wells. The final concentration of α-ketoglutarate in the assay was set to be equivalent to ∼2 × Kmapp.
The human PHD2 assay was performed similarly using the PHD2 protein isoform (the 2X enzyme solution contained 4.0 nM PHD2, 0.2 mg/ml BSA, 400 μM ascorbic acid, 1.25 μg/ml FeCl2, 0.3 mM CHAPS, and 50 mM KCl in 50 mM HEPES at pH 7.5) and HIF1α protein (the 2X substrate solution contained 50 nM Cy5-labeled HIF1α-CODDD, 50 nM VBC, 1 μg/ml streptavidin-Eu, 1 μM α-ketoglutarate, 0.3 mM CHAPS, and 50 mM KCl in 50 mM HEPES at pH 7.5).
The human PHD1 assay was performed similarly using the PHD1 protein isoform (the 2X enzyme solution contained 10 nM PHD1, 0.2 mg/ml BSA, 400 μM ascorbic acid, 1.25 μg/ml FeCl2, 0.3 mM CHAPS, 1 mM tris(2-carboxyethyl)phosphine hydrochloride [TCEP], and 50 mM KCl in 50 mM HEPES at pH 7.5) and HIF1α protein (the 2X substrate solution contained 50 nM Cy5-labeled HIF1α-CODDD, 50 nM VBC, 1 μg/ml streptavidin-Eu, 0.3 μM α-ketoglutarate, 0.3 mM CHAPS, and 50 mM KCl in 50 mM HEPES at pH 7.5). Slight changes included the addition of 0.5 mM final TCEP to the reaction, and plates were read on the EnVision (PerkinElmer) after a 40-minute reaction time.
A modified version of the screening assay was used to examine the mode of inhibition for GSK1278863. Using a 384-well plate (no. 3637; Corning Life Sciences, Tewksbury, MA), PHD3/Cofactor mix (2.33X final concentration in 50 mM HEPES, pH 7.5, with 50 mM KCl) was incubated at room temperature for 1 minute with GSK1278863A (10 point, 3-fold serial dilution in 25% DMSO, 7X final concentration) or 25% DMSO. The enzymatic reaction was initiated upon the addition of substrate (2.33X final concentration) resulting in a final volume of 7 μl. The LANCE ratio values were measured using a PerkinElmer EnVision (excitation at 320 nm, emission at 665 and 615 nm). Final concentrations of the PHD3/Cofactor mix consisted of 5 nM enzyme, 0.1 mg/ml BSA, 5 μM FeCl2, 200 μM ascorbic acid, 70 nM biotin-VBC, 0.6 μg/ml streptavidin-Eu, and 43 nM Cy5-HIF-2α CODDD. The top dose of GSK1278863A tested was 12.5 μM. The substrate mix consisted of a 5 point, 3-fold serial dilution in ddH2O of α-ketoglutarate, where the top concentration tested was 200 μM.
The percentage inhibition data were determined from the slope of the progress curve, typically ranging from 5 to 40 minutes. The IC50 was determined by fitting the percentage inhibition data to a two-parameter IC50 equation. GSK1278863 IC50 values were then plotted as a function of α-ketoglutarate concentration divided by its Kmapp value to determine the inhibition pattern and calculate Kiapp. The data are best fit by the Cheng and Prusoff (1973) equation for competitive inhibition:(1)A second testing under similar conditions resulted in the same conclusion. PHD2 testing consisted of a 5-minute preincubation with GSK1278863A at a top dose of 80 μM. The final enzyme concentration of PHD2 was 2 nM, Cy5-HIF-1α CODDD (60 nM) was used in place of Cy5-HIF-2α CODDD, and the top α-ketoglutarate concentration tested was 6 μM.
Measurement of the intrinsic dissociation rate constant (koff) of GSK1278863A from PHDs was determined by rapid dilution. For part 1a, PHD3/Cofactor mix (1.33X final concentration in 50 mM HEPES, pH 7.5, with 50 mM KCl) was incubated for 30 minutes with DMSO or 10X inhibitor (4X final concentration diluted in 50% DMSO) at room temperature in a deep-well 96-well plate. After a 30-minute incubation, the reactions were diluted 100-fold with buffer plus cofactors (1.11X final concentration). The concentrations of PHD3 and 10X inhibitor were 0.6 and 2.5 μM, respectively, during the incubation and subsequently diluted to 6 and 25 nM.
For part 1b, PHD3/Cofactor mix was diluted 100-fold (6 nM final) and then incubated with 10X inhibitor (2.5 μM) for 30 minutes. In part 2, 9 μl of the diluted incubation mixes from part 1a and 1b were added to a 384-well plate (no. 3637; Corning Life Sciences) containing 1 μl of substrate mix (10X final concentration in 50 mM HEPES, pH 7.5 with 50 mM KCl) with or without 0.1X inhibitor. The LANCE ratio values were measured using a PerkinElmer Envision (excitation at 320 nm, emission at 665 and 615 nm).
Final concentrations after part 2 of the reaction consisted of 5.4 nM PHD3, 0.09 mg/ml BSA, 4.5 μM FeCl2, and 180 μM ascorbic acid for the PHD3/Cofactor mix. GSK1278863A was tested at final concentrations of 2250 nM (10X) and 22.5 nM (0.1X). The substrate mix consisted of 35 nM biotin-VBC, 0.6 μg/ml streptavidin-Eu, 35 nM Cy5-HIF-2a CODDD, and 60 μM α-ketoglutarate. Data are reported as n = 4 determinations, n = 3 used the prior conditions; whereas n = 1 was tested using 450–4.5 nM PHD3 and GSK1278863A where α-ketoglutarate was held at 500 mM.
The half-life was determined by fitting the progress curve to a fixed steady-state velocity equation for time-dependent inhibition.(2)PHD2 testing contained slight experimental adjustments to the PHD3 protocol. From part 1a, a portion of the PHD2/DMSO sample was diluted with 100-fold dilution buffer containing 1) 0X inhibitor, 2) 0.1X inhibitor, or 3) 10X inhibitor while the PHD2/10X inhibitor sample was diluted with buffer containing 0X inhibitor only. Steps part 1b and the addition of 0.1X inhibitor to the substrate mix in part 2 were removed. The final concentrations after part 2 were 1.08 nM PHD2, 900 nM α-ketoglutarate, and 35 nM Cy5-HIF-1α CODDD versus HIF-2α. GSK1278863A was tested at final concentrations of 112.5 (10X) and 1.13 nM (0.1X).
Factor-Inhibiting HIF.
GSK1278863 (10 point, 3-fold serial dilution in 50% DMSO, 15X final concentration) or 50% DMSO was added to 1.1 ml MicroTubes in strips of 12 followed by the addition of human FIH/Cofactor mix (4X final concentration in 50 mM HEPES, pH 7.5, with 50 mM KCl). The enzyme plus inhibitor mix was incubated at room temperature for 30 minutes. Substrate mix (1.4X final concentration in 50 mM HEPES, pH 7.5, with 50 mM KCl) was added to the MicroTubes yielding a 32 μl final reaction volume. Filter paper cut in strips of 12 and saturated with 30 mM Ca(OH)2 were quickly added to the MicroTubes (without touching the reaction mix at the bottom of the tube) and capped. After 25 minutes, the filter paper was removed, placed in a vial with scintillation cocktail, and the released 14CO2 captured on the filter paper was counted in a liquid scintillation spectrometer. The final concentrations of the FIH/Cofactor mix were 50 nM enzyme, 0.25 mg/ml BSA, 10 μM FeCl2, and 1 mM ascorbic acid. The substrate mix consisted of 200 μM HIF-1α (D788-L822), 10 μM α-keto[1-14C] glutarate, and 50 μM α-ketoglutarate. The final concentration of α-ketoglutarate in the assay was set to be equivalent to ∼2 × Kmapp. The percentage inhibition data were determined from the total counts after a 25-minute reaction. The IC50 value was determined by fitting the percentage inhibition data to a two-parameter IC50 equation.
The ability of GSK1278863 to inhibit human FIH in a time-dependent manner was determined using the previously described experimental procedure in addition to an IC50 value determined after 1 minute of enzyme plus inhibitor incubation.
Collagen Prolyl Hydroxylase.
Human CP4H/Cofactor mix (4X final concentration in 50 mM HEPES, pH 7.5, with 50 mM KCl) was incubated at room temperature for 30 minutes with GSK1278863 (10 point, 3-fold serial dilution in 50% DMSO, 15X final concentration) or 50% DMSO. Substrate mix (1.4X final concentration in 50 mM HEPES, pH 7.5, with 50 mM KCl) was added to the plate yielding a 32 μl final reaction volume. After 20 minutes, the reaction was quenched by the addition of 0.67% final H3PO4. Product formation was detected using reverse-phase high-performance liquid chromatography (Ex 336 nM and Em 517 nM). The final concentrations of the CP4H/Cofactor mix were 21.11 nM enzyme, 1 mg/ml BSA, 50 μM FeCl2, 2 mM ascorbic acid, 100 μM dithiothreitol (DTT), and 100 μg/ml catalase. The substrate mix consisted of 100 μM dansyl-GFPG-Oet and 110 μM α-ketoglutarate. The percentage inhibition data were determined from the ratio of product to substrate after a 20-minute reaction. The IC50 value was determined by fitting the percentage inhibition data to a two-parameter IC50 equation.
Inhibition by GSK1278863 against human CP4H was confirmed using the radioactive 14CO2 release assay similar to that described for FIH; however, the reaction was stopped after 20 minutes. The final concentrations of the CP4H/Cofactor mix were 20 nM enzyme, 1 mg/ml BSA, 50 μM FeCl2, 2 mM ascorbic acid, 100 μM dithiothreitol (DTT), and 100 μg/ml catalase. The substrate mix consisted of 100 μM (P-P-G)10, 20 μM α-keto[1-14C] glutarate and 90 μM α-ketoglutarate. The final concentration of α-ketoglutarate in the assay was set to be equivalent to ∼2X Kmapp.
Serotonin (5HT)-2 Receptors.
Because 5-hydroxytryptamine 2B (5HT-2B) receptor agonism has been implicated in the pathogenesis of fenfluramine-induced valvulopathy in humans, we hypothesized it could be involved in the valve lesions we observed. SHSY5Y cells stably transfected with the appropriate 5HT2 receptor were plated in 384-well Nunc noncoated black-walled, clear-bottom plates with 16,000 cells per well in 100 μl α minimum essential medium + ribonucleosides (Gibco Invitrogen, Grand Island, NY) supplemented with 10% fetal bovine serum (Gibco Invitrogen) and 400 μg/ml Geneticin G418 (Gibco Invitrogen) and incubated overnight in a humidified incubator at 37°C and 5% CO2. The cells were washed to remove all media. We added 50 μl of buffer (Hanks’ balanced salt solution, 20 mM HEPES, 4.16 mM NaHCO3, 2.5 mM probenecid, pH 7.4; Sigma-Aldrich, St. Louis, MO) and then aspirated all but 10 μl. We added 30 μl of loading dye (250 μM Brilliant Black, 2 μM Fluo-4 diluted in Hanks’ balanced salt solution buffer; Molecular Devices, Sunnyvale, CA), and the cells were returned to the incubator for 1 hour. An 11-point 1:4 dilution series of test compounds was then added to the cells using the fluorimetric imaging plate reader (FLIPR; Molecular Devices), and the plates were returned to the incubator to equilibrate for 30 minutes.
For agonist assays, 5HT was diluted in buffer at 0.1 mM and then diluted 4-fold in buffer to generate an 11-point serial dilution. Six microliters of the dilution series was transferred into 95 μl buffer. A signal test was performed on a yellow plate to check uniformity of the laser beam, with exposure length set to 0.05 seconds and varying light in the range of 0.5–0.8 W to obtain an average signal of ∼63,000 fluorescence intensity units. A signal test was performed on the cell plate to check variability, with the exposure length set to 0.4 seconds and varying light in the range of 0.5–0.9 W to obtain a desirable signal of ∼6000–10,000 fluorescence intensity units. The following program was then run on the FLIPR: 0.4 second exposure length; read every second for 60 counts for the first sequence, first interval; read every 3 seconds for 20 counts for the second interval; 10 μl addition of fluid. The EC50 values were determined from a four-parameter logistic curve fit using ActivityBase (IDBS, Boston, MA).
For antagonist assays, the following program was run on the FLIPR: 0.4 second exposure length; read every second for 60 counts for the first sequence, first interval; read every 3 seconds for 20 counts for the second interval; 10 μl addition of fluid. The IC50 values were determined from a four-parameter logistic curve fit using ABase.
Mouse Plasma EPO and Hematologic Counts.
B6D2F1/Crl mice were ordered from Charles River Laboratories (Raleigh, NC).
EPO and VEGF.
Normal female B6D2F-1 mice (n = 6) were dosed by oral gavage with GSK1278863 at 60 mg/kg, and blood was collected from euthanized animals by cardiac puncture into EDTA tubes. Vehicle-treated animals were sampled at 6 hours only. Platelet-poor plasma (PPP) was prepared for EPO and VEGF enzyme-linked immunosorbent assay (ELISA) time course analysis as follows. The samples were centrifuged at 1000g for 10 minutes, and the plasma fraction was removed to a new tube. The plasma samples were further centrifuged at 10,000g for 10 minutes. The supernatant (PPP) was removed to a fresh tube and stored at −80°C. The PPP samples were analyzed for EPO and VEGF protein concentrations using Meso-Scale Discovery mouse ELISA plates (MSD, Rockville, MD) according to the manufacturer’s instructions and were read on the MSD SECTOR Imager 6000.
Data from the MSD SECTOR Imager 6000 (n = 1 well for each mouse) were imported into an Excel spreadsheet for analysis. Sample EPO and VEGF concentrations in pg/ml were determined from a regression fit of the EPO or VEGF standard curves. Averages and standard errors were calculated and plotted.
Blood Counts – Reticulocytes and Hemoglobin.
Groups of five female B6D2F1 mice were dosed orally once per day with either vehicle (1% methylcellulose) or GSK1278863 at 3, 10 or 30 mg/kg. Animals were dosed for eight consecutive days, and blood was collected from euthanized animals by cardiac puncture on day 9 into EDTA tubes. Samples were analyzed on the Advia blood analyzer (Siemens Healthcare GmbH, Erlangen, Germany) to determine blood cell parameters.
Blood parameters were imported from the Advia 120 blood analyzer into an Excel spreadsheet (Microsoft, Redmond, WA) for analysis using Hemacalc, an Excel macro. The mean, S.D., and S.E. were calculated using standard formulas. P values were determined using Student’s t test on the square roots of the sample values. The S.E. from Hemacalc was used to display error bars on the graphs.
Screening 14-Day Oral Toxicity Studies
GSK1278863.
Groups of male or female Sprague-Dawley rats (4 per group; obtained from Charles River Laboratories) were given 0 (vehicle), or 10, 30, 60, 100, 250, or 500 mg/kg per day GSK1278863 in 1% methylcellulose once daily for up to 14 days by oral gavage.
Other PHD Inhibitors (Compounds A, B, C, D, E, and F).
Four groups/compound of male Sprague-Dawley rats (4 per group) were given 0 (vehicle), or three dose levels per compound (ranging from 10 to 300 mg/kg per day) in 1% methylcellulose once daily for up to 14 days by oral gavage.
28-Day Oral Toxicity Studies (Compound A).
Five groups of male and female CD-1 mice (10/sex/group; obtained from Charles River Laboratories) were given 0 (vehicle), or four dose levels (ranging from 3 to 100 mg/kg per day) in 1% methylcellulose once daily for up to 28 days by oral gavage.
Four groups of male and female beagle dogs (3/sex/group; obtained from Marshall BioResources, North Rose, NY) were given 0 (empty gelatin capsule), or three dose levels (ranging from 15 to 120 mg/kg per day) once daily for up to 28 days by oral capsule.
The toxicity studies included standard study end points: clinical observations, body weights, hematology and clinical chemistry, organ weights, macroscopic and microscopic observations, and toxicokinetics (performed on separate subsets of animals in the rodent studies).
Results
Identification and Biochemical Characterization of GSK1278863
A novel series of pyrimidinetrione inhibitors was designed to potently inhibit the PHDs, forming tight-binding interactions with the catalytic iron and a cofactor-pocket arginine residue (Fig. 2). GSK1278863 was ultimately identified after extensive optimization of potency, selectivity, and drug metabolism and pharmacokinetics properties (as described further below).
A high-throughput homogeneous fluorescence resonance energy transfer (FRET) assay was developed to measure human PHD enzymatic activity (Pappalardi et al., 2008, 2011). The primary substrate is the HIF CODDD (Fig. 3), which contains a proline residue hydroxylated by PHD. All biochemical studies conducted with PHD1 and PHD2 used HIF1α CODDD as the substrate, and PHD3 used HIF2α CODDD. The HIF CODDD substrates are labeled with the fluorescent group Cy5. When hydroxylated, Cy5-HIF CODDD is recognized by the biotin-labeled VBC complex. Addition of streptavidin-Eu chelate results in proximity of Eu to Cy5 in the product, allowing for detection by FRET. A ratio of Cy5 to Eu fluorescence emission, the lanthanide chelate excite (LANCE) ratio is the ultimate readout, as this normalized parameter has significantly less variance than the Cy5 emission alone.
The inhibitory mechanism of GSK1278863 against α-KG for PHD3 was examined by plotting GSK1278863 IC50 values as a function of α-KG concentration divided by its Kmapp value. GSK1278863 IC50 values increased in a manner consistent with an α-KG-competitive mode of inhibition, where the data are best fit by eq. 1, the Cheng and Prusoff (1973) equation for competitive inhibition, yielding a Kiapp value of 92 nM for PHD3 (Fig. 4, left panel).
GSK1278863 demonstrated time-dependent inhibition against PHD3, as evidenced by a large increase in potency after a 30-minute enzyme:inhibitor preincubation, shifting the Kiapp from 92 nM (1 minute enzyme:inhibitor preincubation) to 1.8 nM (Table 1). The Ki*app values of 1.8 and 7.3 nM calculated using the Cheng-Prusoff equation for competitive inhibition against PHD3 and PHD2, respectively, may underestimate the true Ki (Ki*) under these conditions if the EI* conformation was not fully obtained under the standard assay conditions which contained a 30-minute enzyme:inhibitor preincubation (Table 1).
To determine whether the observed time-dependent behavior was the result of slow-binding reversible inhibition between GSK1278863 and PHD3, a preformed GSK1278863–PHD3 complex was rapidly diluted into substrate mix where the α-KG concentration was roughly 6 × Kmapp to prevent rebinding (Copeland, 2005). Upon dilution, recovery of PHD3 activity was observed, indicating that GSK1278863 is a reversible inhibitor of PHD3.
The observed rate of recovery, kobs, was obtained by fitting the data to a single exponential model. Under experimental conditions, kobs reflects the intrinsic dissociation rate (koff) of the inhibitor from PHD3. The residence half-life value (t1/2) is calculated from koff based on the equation t1/2 = 0.693/koff; t1/2 = 132 ± 19 minutes (n = 4) for PHD3 (Fig. 4, right panel). This t1/2 value indicates that GSK1278863 does not dissociate instantaneously from the PHD3 enzyme, resulting in prolonged inhibition of PHD3, which may impact the in vivo pharmacologic activity of this compound. GSK1278863 was also found to be an α-KG competitive, reversible inhibitor of PHD2 with a residence half-life, t1/2 = 92 ± 43 minutes (n = 2) (see Supplemental Data).
Because GSK1278863 was designed as a cofactor mimetic, its ability to inhibit additional members of the iron-dependent α-KG- dioxygenase superfamily was assessed. The IC50 values were determined against human CP4H using a procollagen peptide substrate and human FIH using a HIF1α peptide (D788-L822) after 30 minutes of enzyme plus inhibitor preincubation. The resulting IC50 values of 9.8 µM for FIH and >200 µM for CP4H show that GSK1278863 is at least 1000-fold selective for PHDs (Table 1).
The Biology and Pharmacology of GSK1278863
Induction of EPO.
The immediate downstream effect of PHD inhibition in a cellular context is HIFα subunit accumulation. Stabilization of HIF1α and HIF2α was determined by Western blot analysis of nuclear protein extracts after GSK1278863 treatment of Hep3B cells. Western blot analysis demonstrated that neither HIF1α nor HIF2α was detected in the vehicle-treated cells, and both HIF1α and HIF2α were visualized in the DFX-positive control-treated cells. Treatment with either 25 or 50 μM GSK1278863 for 6 hours resulted in the accumulation of both HIF1α and HIF2α subunits (Fig. 5). These results demonstrate that prolyl hydroxylase inhibition by GSK1278863 treatment of cells results in the immediate downstream effect of HIFα subunit stabilization.
Moving from an in vitro context to in vivo context, normal female B6D2F1 mice were administered a single oral dose of GSK1278863 at 60 mg/kg, and blood samples were collected at intervals between 4 and 30 hours after dosing (n = 6 mice/time point for GSK1278863-treated mice; vehicle-treated mice were sampled at 6 hours only). After the treatment with GSK1278863, the EPO protein levels peaked at 12 hours after dosing (Fig. 6), representing an 11.2-fold increase with a mean plasma concentration of 1303 pg/ml. Additionally, the EPO values at all other time points remained elevated by 1.9- to 2.9-fold relative to the vehicle-treated mice. The VEGF concentrations remained generally unchanged across the time course and were only slightly higher than those of vehicle-treated mice. These data indicate that a single 60 mg/kg dose of GSK1278863 results in a significant but transient increase in circulating levels of EPO, with minimal impact on VEGF concentrations.
In Vivo Increases in Reticulocyte Count and Hemoglobin.
The pharmacologic consequence of repeated daily treatment of normal B6D2F1 mice with vehicle or GSK1278863 (3, 10, or 30 mg/kg per day) for 8 days is shown in Fig. 7. Approximately 24 hours after the final dose, the mice were euthanized, and their blood was collected by cardiac puncture and analyzed on the Advia Blood Analyzer. Statistically significant increases in reticulocytes were observed in the compound-treated mice (211%, 544%, and 673% over placebo) administered 3, 10, and 30 mg/kg GSK1278863, respectively (Fig. 7A).
Consistent with the observed reticulocyte increases, the hemoglobin levels were also found to be significantly elevated (12%, 17%, and 13%) relative to vehicle-treated mice after dosing with 3, 10, and 30 mg/kg per day GSK1278863, respectively (Fig. 7B). Red blood cell counts and hematocrit were statistically significantly elevated in mice as well. Similar responses in reticulocytes and red blood cell mass parameters were evident in rats, beagle dogs, and cynomolgus monkeys after comparable treatment durations, with responses progressing as dosing duration became more chronic (data not shown).
Pharmacokinetics of GSK1278863 in Different Animal Species
When dosed intravenously, GSK1278863 showed low blood clearance in the mouse, rat, dog, and monkey [<1% hepatic blood flow (Qh) in mouse and rat, ∼3% Qh in dog, and ∼19% Qh in monkey] (Table 2). The steady-state volume of distribution (∼0.3–0.8 l/kg) was less than or approximately equal to the total body water in all four preclinical species.
Pharmacokinetic parameters were also determined for GSK1278863 after oral suspension administration to rats (Table 3). Unless otherwise indicated, the rats were fasted overnight before oral administration of GSK1278863. The systemic exposure of GSK1278863 was similar in the rat after oral solution dosing (1.6 mg/kg) and oral suspension dosing (31.1 mg/kg) of crystalline GSK1278863, suggesting a minimal impact of solubility and dissolution rate on the bioavailability of the drug molecule. Similar favorable comparisons between orally dosing a solution versus a suspension of crystalline solid were observed in both mice and dogs (see Supplemental Tables 1 and 2).
Toxicology
In screening 14-day oral toxicity studies with once daily administration to Sprague-Dawley rats, GSK1278863 demonstrated an acceptable toxicity profile that supported candidate selection and continued development. In chronic oral toxicity studies, dose-limiting effects of GSK1278863 were the result of marked pharmacologically mediated increases in hematocrit and consisted of generalized vascular congestion, thrombosis, and/or multiorgan pathology. These findings are considered consequences of compromised blood flow and vascular perfusion consequent to high hematocrits and inferred high blood viscosity, and they represent a manageable clinical risk via the routine hematology monitoring and hemoglobin stopping criteria in place in clinical studies of GSK1278863.
In contrast, four compounds (compounds A, B, C, and D; Chai et al. (2007), Shaw et al. (2008); see Supplemental Fig. 2) of similar chemical structure and comparable on-target potencies induced cardiac valve lesions in oral toxicity studies in 14-day rats or 28-day mice and dogs. These lesions, which were observed in right and/or left atrioventricular valves and/or aortic valves, consisted of a spectrum of changes most notably including myxomatous thickening characterized by widened valve leaflets and/or neutrophilic or mixed inflammatory cell infiltrates (Fig. 8).
These valve lesions were not associated with an attached thrombus or thrombosis elsewhere in the heart, as has been described for the valvulopathy observed in 4-week rat toxicology studies conducted with the hyperglycosylated rHuEPO AMG-114 (Andrews et al., 2014a, Sinclair, 2013). Uncertainty over the relevance of these findings to humans resulted in an unfavorable risk assessment for continued development of these four compounds. Because valvulopathies are commonly associated with several human connective tissue disorders (e.g., Ehlers Danlos, Marfans) (Glesby and Pyeritz, 1989; Ha et al., 1994; Myllyharju and Kivirikko, 2001; Lincoln et al., 2006), which are characterized by mutations in collagens and/or collagen processing genes, it was hypothesized that inhibitory activity on the closely related enzyme CP4H may be contributory. Interestingly, these four compounds demonstrated moderate inhibitory activity on human CP4H (IC50 of 2.5–63 µM; PHD:CP4H selectivity of 20- to 1000-fold).
Prioritizing PHD:CP4H selectivity in our compound selection criteria led to the discovery of GSK1278863 (human CP4H IC50 >200 µM; PHD:CP4H selectivity >9000-fold), which did not cause valvulopathy in 14- or 28-day rat oral toxicity studies or in dog or monkey studies of up to 9 months duration when dosed to maximum tolerated doses. In 3- and/or 6-month mouse and/or rat oral toxicity studies, a few GSK1278863-treated animals presented with thrombi in atria or ventricles that resulted in minimal secondary valve lesions in adjacent structures. Thromboses in these cases were considered complications from polycythemia. The resultant valvular changes were morphologically distinct in character and severity from those observed in studies with compounds A through D and were typical of those often observed in association with cardiac thrombosis.
In light of the implication of 5HT-2B receptor agonism in the pathogenesis of fenfluramine-induced valvulopathy in humans (Connolly et al., 1997), we tested GSK1278863 and compounds A through D for antagonism of 5HT-2A, -2B, and -2C receptors and for agonism of 5HT-2B and -2C receptors. Activities were mild (XC50 > 10 μM) and comparable for all compounds, suggesting serotonergic receptor-mediated mechanisms as an unlikely factor.
Importantly, GSK1278863 and two subsequent candidates (compounds E and F; human CP4H IC50s >200 µM) that did not cause valvulopathy in screening 14-day rat oral toxicity studies increased hematocrits to magnitudes comparable to those induced by the four compounds that induced cardiac valve lesions (compounds A through D). This suggests that polycythemia and assumed increased blood viscosity were not significant or the sole contributing factors in the valvulopathy observed in short-term studies, especially given the absence of valvular or cardiac thromboses. Thus, our experience suggests a plausible association between CP4H inhibitory activity and the cardiac valve lesions observed with several nonselective HIF-prolyl-hydroxylase inhibitors in 14- or 28-day dog and/or rodent oral toxicity studies.
Discussion
Our search for novel and potent HIF prolyl hydroxylase inhibitors began using the validated principle of molecular mimicry. N-Oxalylglycine has long been used as an unreactive substrate mimetic to inhibit α-KG-utilizing enzymes such as the prolyl 4-hydroxylases (Fig. 2) (Cunliffe et al., 1992; Mole et al., 2003; Hausinger, 2004). Extending this principle, several heteroaryl-glycinamides have been discovered as inhibitors of α-KG-utilizing dioxygenases using a heterocyclic ring nitrogen atom to form a five-membered chelate with the enzyme-bound iron (e.g., {[(4-hydroxy-8-iodoisoquinolin-3-yl)carbonyl]amino}acetic acid) (McDonough et al., 2006). We postulated that an inhibitor with the potential to form a six-membered chelate would form a strong interaction with the PHDs and result in potent inhibitors of this class of enzyme. As shown in Fig. 2, GSK1278863 (a pyrimidinetrione-glycinamide) mimics the binding of N-oxalylglycine, chelating the catalytic iron through the glycinamide carbonyl and an acidic ring hydroxyl group in a six-membered chelate. The specific ring substitutions on the pyrimidine nucleus of GSK1278863 were derived through an extensive lead-optimization campaign and provided an inhibitor with optimized biologic profile and pharmaceutic properties for progression into clinical trials (Johnson et al., 2014; Brigandi et al., 2016; Holdstock et al., 2016; Akizawa et al., 2017); a full description of this campaign is outside the scope of the present discussion.
GSK1278863 is a novel small molecule agent that stimulates erythropoiesis through inhibition of PHDs. GSK1278863 is a potent inhibitor of all three HIF prolyl hydroxylase isozymes, PHD1, PHD2, and PHD3, with Ki apparents in the single-digit nanomolar range. Inhibition of the HIF-regulatory prolyl hydroxylases causes the accumulation of HIFα transcription factors, resulting in increased transcription of HIF-responsive genes. The downstream physiologic consequences of this altered gene expression essentially simulate the body’s natural adaptive response to hypoxia.
GSK1278863 is highly selective for the HIF-prolyl hydroxylases compared with related oxygen-dependent, α-KG-utilizing metalloenzymes such as FIH and CP4H. GSK1278863 treatment of Hep3B cells induces downstream effects of HIF-prolyl hydroxylase inhibition, including stabilization of HIF1α and HIF2α, and induction of EPO. A single oral dose of GSK1278863 leads to peak plasma EPO concentrations in mice within 12 hours; daily dosing of GSK1278863 in preclinical species (see Supplemental Data) leads to a robust pharmacodynamic response, resulting in a significant increase in hemoglobin and associated erythroid parameters within 7–14 days.
In toxicity studies, GSK1278863 has demonstrated an acceptable safety profile and therapeutic window, supporting clinical progression into phase 3. Importantly, with several precandidate compounds, we observed an association between CP4H inhibitory potency and development of dog and/or rodent cardiac valve lesions. Although further work would be required to investigate this association, it seems plausible as CP4H plays a key role in collagen synthesis and triple-helix formation, which, if impaired, could impact the organization of extracellular matrix.
For example, it has been reported that interference with fibroblast/interstitial cell-mediated matrix turnover can impact the integrity and repair of cardiac valves (Jaffe et al., 1981; Glesby and Pyeritz, 1989; Gamulescu et al., 2006; Hinton et al., 2006; Hakuno et al., 2009). This is exemplified by the observation that an activin receptor-like kinase 5 (Alk5) inhibitor, a potent antifibrosis agent, causes severe and irreversible cardiac valve lesions in rats (Anderton et al., 2011). Also, Andrews et al. (2014a,b) have reported on valvulopathy related to erythropoiesis-stimulating agent in rats administered AMG-114 and several other recombinant human erythropoietins. The AMG-114-related rat valvulopathy was often accompanied by an attached thrombus. This was not the case for the four GSK PHD compounds that induced valvulopathy in short-term studies. The Andrews et al. article suggests increased hematocrit as a predisposing factor in erythropoiesis-stimulating agent–related rat toxicities, including valvulopathy. Our experiences with several PHD inhibitors that caused valvulopathy in dogs and/or rodents point to a plausible association with off-target inhibitory activity on the closely related enzyme CP4H. Importantly, GSK1278863A, the molecule in clinical trials, is >9000-fold selective for PHDs over CP4H.
PHDs and CP4H belong to a growing enzyme superfamily of iron and α-ketoglutarate-dependent dioxygenases. The members of this enzyme superfamily play critical roles in hypoxic signaling, DNA repair, stress response mechanisms, lipid and growth factor metabolism, and epigenetic modifications (Aravind and Koonin, 2001; Falnes et al., 2002; Chen et al., 2008; Yang et al., 2009; Aik et al., 2013; Johansson et al., 2014; Ichiyama et al., 2015). Among them, many are important targets for drug discovery and development. As most of the inhibitors of these enzymes are designed to mimic the binding of N-oxalylglycine involving chelation of the catalytic iron and competing for α-ketoglutarate binding (Joberty et al., 2016), our finding cautions that CP4H inhibition might be an undesirable off-target activity.
In summary, we have described the preclinical development of the PHD inhibitor GSK1278863. It is a potent and selective inhibitor, demonstrating ∼600- to 2800-fold selectivity against FIH and >9000-fold selectivity against CP4H. GSK1278863 stabilized HIF1α and HIF2α in Hep3B cells and led to production of EPO and subsequent increases in reticulocytes and hemoglobin in vivo. The therapeutic window provided by the demonstrated efficacy and acceptable safety margin supported the progression of GSK1278863 (daprodustat) into the clinic, where it has been evaluated in phase I (NCT00750256, NCT01319006, NCT01376232, NCT01673555, NCT02348372) and phase II (NCT01047397), and is in ongoing phase III clinical trials (NCT02876835, NCT02879305) for the treatment of anemia of chronic kidney disease.
Acknowledgments
The authors thank Erding Hu, Kendall Frazier, and Tim Hart for their assistance with study management and critical review, and Tracy Gales for image support, all of whom are employees of GlaxoSmithKline. All listed authors met the criteria for authorship set forth by the International Committee for Medical Journal Editors.
Authorship Contributions
Participated in research design: Adams, Ariazi, Duffy, Erickson-Miller, Luo.
Conducted experiments: Ariazi, Biju, Pappalardi.
Contributed new reagents or analytic tools: Duffy, Fitch, Shaw, Wiggall.
Performed data analysis: Adams, Ariazi, Duffy, Hugger, Luo, Pappalardi.
Wrote or contributed to the writing of the manuscript: Adams, Ariazi, Duffy, Pappalardi.
Footnotes
- Received April 28, 2017.
- Accepted July 26, 2017.
This work was supported by GlaxoSmithKline.
↵This article has supplemental material available at jpet.aspetjournals.org.
Abbreviations
- BSA
- bovine serum albumin
- CHAPS
- 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate
- CODDD
- C-terminal oxygen-dependent degradation domain
- CP4H
- collagen prolyl hydroxylase
- DFX
- deferoxamine
- DMSO
- dimethylsulfoxide
- ELISA
- enzyme-linked immunosorbent assay
- EPO
- erythropoietin
- Eu
- Europium
- FIH
- factor-inhibiting HIF
- FLIPR
- fluorimetric imaging plate reader
- FRET
- fluorescence resonance energy transfer
- GSK1278863
- 2-[(1,3-dicyclohexyl-2,4,6-trioxo-1,3-diazinane-5-carbonyl)amino]acetic acid (daprodustat)
- HIF
- hypoxia-inducible factor
- 5HT
- 5-hydroxytryptamine
- α-KG
- α-ketoglutarate
- PHD
- hypoxia-inducible factor prolyl hydroxylase
- PPP
- platelet-poor plasma
- rHuEPO
- recombinant human erythropoietin
- TBST
- Tris-buffered saline containing 0.1% Tween-20
- TCEP
- tris(2-carboxyethyl)phosphine hydrochloride
- VBC
- von Hippel-Lindau protein-elongin B-elongin C
- VEGF
- vascular endothelial growth factor
- Copyright © 2017 by The American Society for Pharmacology and Experimental Therapeutics