Selexipag Active Metabolite ACT-333679 Displays Strong Anticontractile and Antiremodeling Effects but Low β-Arrestin Recruitment and Desensitization Potential

Prostacyclin (PGI2) receptor (IP receptor) agonists, which are indicated for the treatment of pulmonary arterial hypertension (PAH), increase cytosolic cAMP levels and thereby inhibit pulmonary vasoconstriction, pulmonary arterial smooth muscle cell (PASMC) proliferation, and extracellular matrix synthesis. Selexipag (Uptravi, 2-{4-[(5,6-diphenylpyrazin-2-yl)(isopropyl)amino]butoxy}-N-(methylsulfonyl)acetamide) is the first nonprostanoid IP receptor agonist, it is available orally and was recently approved for the treatment of PAH. In this study we show that the active metabolite of selexipag and the main contributor to clinical efficacy ACT-333679 (previously known as MRE-269) behaved as a full agonist in multiple PAH-relevant receptor-distal—or downstream—cellular assays with a maximal efficacy (Emax) comparable to that of the prototypic PGI2 analog iloprost. In PASMC, ACT-333679 potently induced cellular relaxation (EC50 4.3 nM) and inhibited cell proliferation (IC50 4.0 nM) as well as extracellular matrix synthesis (IC50 8.3 nM). In contrast, ACT-333679 displayed partial agonism in receptor-proximal—or upstream—cAMP accumulation assays (Emax 56%) when compared with iloprost and the PGI2 analogs beraprost and treprostinil (Emax ∼100%). Partial agonism of ACT-333679 also resulted in limited β-arrestin recruitment (Emax 40%) and lack of sustained IP receptor internalization, whereas all tested PGI2 analogs behaved as full agonists in these desensitization-related assays. In line with these in vitro findings, selexipag, but not treprostinil, displayed sustained efficacy in rat models of pulmonary and systemic hypertension. Thus, the partial agonism of ACT-333679 allows for full efficacy in amplified receptor-distal PAH-relevant readouts while causing limited activity in desensitization-related receptor-proximal readouts.


Introduction
Prostacyclin (PGI 2 ) is an arachidonic acid metabolite synthesized mainly through endothelial cells by cyclooxygenase enzymes in conjunction with PGI 2 synthase. PGI 2 serves as an autocrine/paracrine mediator, modulating various physiologic processes by binding and activating the prostacyclin I 2 receptor (IP receptor), a G-protein coupled receptor (GPCR) that stimulates adenylate cyclase to synthesize cAMP (Boie et al., 1994). The IP receptor is highly expressed in cell types such as vascular smooth muscle cells, fibroblasts, platelets, and leukocytes. Insufficient PGI 2 synthesis and IP receptor signaling have been linked to cardiovascular pathologies, especially pulmonary arterial hypertension (PAH) (Christman et al., 1992;Tuder et al., 2001;Arehart et al., 2008). PAH is a rare disease characterized by increased pressure in the pulmonary circulation caused by constriction and progressive remodeling of the pulmonary vasculature and thus increased pulmonary vascular resistance (Galie et al., 2009;Morrell et al., 2009). Exogenous supplementation of synthetic PGI 2 (epoprostenol) by continuous i.v. application was the first efficacious therapy in PAH (Barst et al., 1996).
Currently, three PGI 2 analogs have been marketed for treatment of PAH: iloprost, treprostinil, and beraprost (Olschewski et al., 2004). Several limitations of PGI 2 analog therapy are known. They are not selective-that is, they activate other prostanoid receptors-and they display chemical and metabolic instability limiting their potential for oral administration (Kuwano et al., 2007;Whittle et al., 2012). In addition, being GPCR agonists, it has been suggested that PGI 2 analogs might have the potential to cause tachyphylaxis (Rubin et al., 1990;Archer et al., 1996;Shapiro et al., 1997;MacLaughlin et al., 1998;Barst et al., 2003), making it necessary to escalate drug dosing over time. Tachyphylaxis can be caused by different mechanisms. In the case of PGI 2 and its analogs, one of the main mechanisms is thought to be classic receptor internalization into the endocytic compartment (Smyth et al., 2000(Smyth et al., , 2002Hasse et al., 2003;O'Keeffe et al., 2008).
To overcome the aforementioned pharmacologic limitations associated with PGI 2 structures, Actelion/Nippon Shinyaku have developed a selective, oral nonprostanoid IP receptor agonist, selexipag (2-{4-[(5,6-diphenylpyrazin-2-yl)(isopropyl) amino]butoxy}-N-(methylsulfonyl)acetamide, previously known as NS-304) (Asaki et al., 2015), which has been recently approved in the United States, the European Union, and Japan for the treatment of PAH (Sitbon et al., 2015). After systemic absorption, selexipag, the parent drug, is hydrolyzed to the highly potent metabolite (isopropyl)amino]butoxy}acetic acid, previously known as MRE-269), the major driver of clinical efficacy ( Fig. 1) (Kuwano et al., 2007(Kuwano et al., , 2008. ACT-333679 differentiates favorably from PGI 2 analogs in terms of its higher metabolic stability and its high IP receptor selectivity. To date, the in vitro activities of selexipag and in particular its active metabolite ACT-333679 using PAH-relevant assays in pulmonary arterial smooth muscle cells (PASMC) have not been reported. Furthermore, it is currently not known how ACT-333679 compares with PGI 2 analogs regarding its IP receptor internalizing and desensitizing properties.
In the present study we compared the activity of ACT-333679 with that of the prototypic PGI 2 analog iloprost by measuring diverse PAH-relevant parameters in PASMC such as cellular relaxation, proliferation, and extracellular matrix synthesis and found both compounds to be full agonists. We then uncovered the unique partial agonism of ACT-333679 in receptor proximal readouts such as induction of cAMP synthesis, recruitment of b-arrestin, and IP receptor internalization. This partial agonism is in contrast to the full agonism displayed by iloprost and other PGI 2 analogs (beraprost, treprostinil). Finally, the propensity of IP receptor agonists to induce tachyphylaxis was tested in vivo comparing selexipag with the PGI 2 analog treprostinil. Our data suggest that the unique partial agonism exhibited by ACT-333679 on receptorproximal parameters reduces its potential for desensitization while still reaching maximal effects on downstream parameters relevant to PAH.
The human IP receptor-expressing cell pool T-REx-HEK-293-CMV. TO.FLAG-humanIP-C9 (abbreviated T-REx-HEK-hIP [HEK, human embryonic kidney cells]) was generated by integrase-mediated homologous recombination of the IP receptor sequence into the T-REx HEK293 background. This expression system allows the tetracycline-inducible expression of genes of interest from one defined insertion site, and the generated cells were used 1) in cAMP experiments to evaluate cAMP responses to IP receptor agonists at high and low expression levels and 2) in flow cytometry to observe agonist-induced IP receptor internalization. Parental T-REx-HEK293 cells (Life Technologies) were cultivated in growth medium: Dulbecco's modified Eagle's medium 1 GlutaMAX-I (Life Technologies), 10% dialyzed FBS (Life Technologies), 100 U/ml penicillin, 100 mg/ml streptomycin, 0.1 mg/ml hygromycin B (Life Technologies), 5 mg/ml blasticidin (Life Technologies), and 1% nonessential amino acid solution (Life Technologies). Recombinant T-REx-HEK-hIP cells were cultivated in selection medium (growth medium containing 1 mg/ml Geneticin [Life Technologies] instead of hygromycin). Every second day the medium was exchanged with fresh selection medium.
Before stimulation with IP receptor agonists, differentiation, or growth factors, the cells were brought into the quiescent state by cultivation in starvation medium: SmBM, 20 mM HEPES, 100 mg/ml penicillin/streptomycin (Life Technologies), and 0.1% fatty acid-free bovine serum albumin (BSA; Calbiochem, San Diego, CA). Cells were used up to passage 4 and were harvested for seeding for the various assays at approximately 90% cell density.
Impedance Assays. PASMC were seeded into E-plates (ACEA Biosciences, San Diego, CA) at 5000 cells/well and placed into the xCELLigance MP device (ACEA Biosciences). The next day, the medium was exchanged for differentiation medium. After 48 hours, the cells were starved for 6 hours in starvation medium and then subjected to IP receptor agonists diluted in starvation medium and further cultivated for several hours.
Impedance measurements were performed during the whole experimental period. For analysis, impedance traces were aligned at the last time point before the IP receptor agonist addition, the vehicle baseline was subtracted, and the impedance minima within the first 3 hours after agonist addition were used to generate concentration-response curves. EC 50 and maximal efficacy (E max ) values were calculated with the proprietary IC 50 -Witch software (Actelion Pharmaceuticals, Allschwil, Switzerland) using the compound intrinsic curve maximum and minimum as plateau values. Compound efficacies were compared with the maximal efficacy of iloprost (100%).

Efficacy and Desensitization Profile of Selexipag
For the endothelin-1 (ET-1)/IP receptor agonist combination experiments, the cells were prepared as previously described, and after the starvation period they were stimulated with 10 nM ET-1 (diluted in starvation medium), observed for 60 minutes until a response plateau was reached, and then stimulated with a dilution series of ACT-333679 or iloprost in starvation medium, again followed by an observation period.
For analysis, impedance traces were aligned at the last time point before ET-1 addition, the vehicle-vehicle baseline was subtracted, and the impedance minimum 60 minutes after the IP receptor agonist addition was used to generate concentration-response curves.
Tomographic Microscopy. Human PASMC were seeded into 35-mm tissue culture dishes (FluoroDish; World Precision Instruments, Sarasota, FL) at 15,000 cells/dish, grown overnight, and then treated for 24 hours in differentiation medium. Then cells were starved for 3 hours by exchange for starvation medium with 20 mM HEPES. The dishes were then placed on the tomographic microscope (3D Cell Explorer, Nanolive S.A., Lausanne, Switzerland) and threedimensional tomographic images (z-stacks) were taken at regular intervals (2 minutes).
Within the observation period, cells were treated with ET-1 (100 nM) for ∼90 minutes followed by ACT-333679 (1 mM) or vehicle. The plane with the best focus was picked from every z-stack and was used to generate time lapse sequences. To get a smooth motion movie, the number of frames was increased by a combination of motion interpolation and frame blending. In brief, a similarity distance between pixels in different frames was defined by making use of the intensities of surrounding pixels. A motion vector was defined from the position of each pixel in frame "n" to the closest pixel in frame "n 1 1," based on the similarity distance. This vector allowed a smooth interpolation which was used to increase the frame rate from 0.5 to 60 frames per second. Further postprocessing steps such as noise reduction, gamma, and contrast enhancement were applied to the final rendering of the video.
Immunoblotting. Isogenic T-REx-HEK-hIP cells were seeded at 15,000 cells/well into 96-well plates and were supplemented after attachment with different concentrations of tetracycline between 0 and 10 ng/ml to induce increasing IP receptor expression levels. After a 24-hour induction period, the cells were washed with phosphate-buffered saline (PBS) and then lysed on ice in 15 ml/well of RIPA buffer (Sigma-Aldrich, Buchs, Switzerland) containing 100 mM NaF, 4 mM Na-orthovanadate, 1 mM phenylmethylsulfonylfluoride (PMSF), 1 mM dithiothreitol (DTT), and 100 U/ml benzonase (Sigma-Aldrich).
The contents of one well (15 ml volume containing 10 mg of protein) were analyzed by SDS-PAGE and Western blotting using mouseanti-human IP receptor monoclonal antibody 13-Q (1 mg/ml; Santa Cruz Biotechnology, Dallas, TX) and horseradish peroxidase-coupled anti-mouse IgG secondary antibodies (GE Life Sciences, Glattbrugg, Switzerland). Membranes were treated with Western Lightning- Enhanced Chemiluminescence Substrate (PerkinElmer, Schwerzenbach, Switzerland), and the chemiluminescence signal was recorded and quantified using a chemiluminescence reader (LAS-4000; Fujifilm, Tokyo, Japan) and the corresponding software (Multigauge V3.0; Fujifilm).
Human PASMC were seeded at 10,000 cells per well into 96-well plates that were precoated with fibronectin. The next morning the growth medium was exchanged for differentiation medium, and after 48 hours the cells were switched for 6 hours to starvation medium. The cells were then stimulated for 90 minutes at 37°C with dilution series of IP receptor agonists (in starvation medium; for MLCK phosphorylation) or with dilution series of IP receptor agonists for 24 hours at 37°C (in starvation medium; 650 ng/ml platelet-derived growth factor-BB [PDGF-BB]; for cyclin D 1 and p27/Kip1). Fig. 3. Effect of the IP receptor agonist ACT-333679 on cellular shape change in human PASMC as determined by time lapse tomographic microscopy. Cells were treated with ET-1 (100 nM) to induce cell contraction. After 70-90 minutes cells were then either treated with (A) ACT-333679 (1 mM) to induce relaxation or (B) vehicle. Alternatively, cells were (C) treated twice with vehicle. Tomographic images were taken at regular intervals along the whole experimental period (∼180 minutes), and a set of images along the time course is shown with the baseline cell circumference indicated by a dotted line. Bar: 20 mm. Representative sequences of one experiment out of n = 3 experiments. The corresponding movie sequences can be accessed via http:// orbit.actelion.com/selexipag/.
[ 3 H]-Thymidine Incorporation Assays. DNA synthesis as a measure of cell proliferation was determined with the [ 3 H]-thymidine incorporation assay. Human PASMC were seeded at 3,200 cells/well in 96-well plates in 100 ml/well growth medium and cultured overnight at 37°C/5% CO 2 . The plates were sealed (Breathseal; Greiner, Loerrach, Germany) after every experimental manipulation. After 20 hours the cells were cell-cycle arrested in 100 ml/well of starvation medium for 24 hours at 37°C/5% CO 2 . Then the cells were supplemented with a dilution series of IP receptor agonists or a dilution series of DMSO (a vehicle control curve was used on every plate) and incubated for 2 hours at 37°C/5% CO 2 , followed by the addition of 2.5 ng/ml or 10 ng/ml PDGF-BB diluted in PASMC starvation medium and [ 3 H]thymidine (PerkinElmer) diluted in starvation medium (0.6 mCi/w, 10 ml/well). The cells were incubated for 24 hours at 37°C/5% CO 2 . Several control wells per plate were not treated with PDGF-BB for the determination of basal [ 3 H]-thymidine incorporation.
Finally, the medium was removed, and the cells were detached with 100 ml/well of 0.25% trypsin-EDTA solution (Life Technologies) at 37°C. After 10 minutes, cellular DNA was precipitated by the addition of 100 ml/well of 20% (w/v) ice-cold trichloroacetic acid (final concentration 10%) with incubation on ice for 1 hour. The DNA was collected with the cell harvester on glass fiber filters (Unifilter-96, GF/C; PerkinElmer). The filters were dried, the liquid scintillator (Microscint 20; PerkinElmer) was added, and the plate was subjected to liquid-scintillation counting (TopCount; PerkinElmer).
[ 3 H]-Proline Incorporation Assays. The [ 3 H]-proline incorporation assay determined the extracellular matrix synthesis (collagens and fibronectin contain high amounts of proline) as a measure of fibrosis. Human PASMC were seeded at a density of 10,000 cells/well onto fibronectin-coated 96-well plates in 100 ml/well complete growth medium and were cultured for 8-15 hours at 37°C/5% CO 2 . Then medium was exchanged for 100 ml/well starvation medium; after overnight incubation, the cells were supplemented with a dilution series (10 ml/well) of ACT-333679 or iloprost and with 25 or 50 ng/ml human recombinant PDGF-BB (Sigma-Aldrich) or of vehicle. Cells were cultivated at 37°C/5% CO 2 for 40 hours. During the last 24 hours of this incubation period, the wells were supplemented with 10 ml/well of L-[2,3-3 H]-proline in starvation medium (0.2 mCi/well) (PerkinElmer).
For determination of [ 3 H]-proline incorporation, cell supernatants were discarded, the cells were lysed in 150 ml/well NaOH (0.15 M), and the lysates were incubated on ice for 30 minutes. For protein precipitation, 100 ml/well of trichloroacetic acid (50% (w/v) stock, final concentration 20%) was added, and the lysates were incubated on ice for 30 minutes. Precipitated proteins were collected with the Filtermate cell harvester (PerkinElmer) onto glass fiber filters (Unifilter-96, GF/C; PerkinElmer); the filters were washed 8 times with deionized water, dried, and then supplemented with 60 ml/well of liquid scintillator (Microscint20; PerkinElmer). Then the plates were subjected to liquid scintillation counting (TopCount; PerkinElmer).
Cyclic AMP Assay. T-REx-HEK-hIP were seeded at 20,000 cells/well into 96-well plates, and grown overnight in presence of 1 ng/ml or 10 ng/ml tetracycline (Sigma-Aldrich) to induce low or high IP receptor expression levels, respectively. Cells were washed with assay buffer (1x Hank's balanced salt solution: 20 mM HEPES,   Gatfield et al. 0.0375% NaHCO 3 , 0.1% fatty acid-free BSA) and then subjected to an agonist dilution series in assay buffer in the presence of 3-isobutyl-1methylxanthine (0.5 mM; Sigma-Aldrich).
After 30 minutes, the cells were lysed, the cAMP levels were determined using the Tropix cAMP-screen System (ThermoFisher Scientific, Reinach, Switzerland), and the luminescence was read with a Synergy4 microplate reader (BioTek Instruments, Winooski, VT). Data were converted into concentration-response curves, and the EC 50 values and E max values were calculated with the proprietary IC 50 -Witch software (Actelion Pharmaceuticals) using the compound intrinsic maximum and minimum as plateau values. Compound E max values were compared with the maximal efficacy of iloprost (100%).
For PASMC, a slightly modified protocol was used: 10,000 cells/well were seeded and cultivated for 48 hours in differentiation medium and 6 hours in starvation medium before stimulation with IP receptor agonists for 30 minutes.
b-Arrestin Recruitment Assay. CHO-hIP PathHunter cells were detached, seeded in 384-well plates, and grown overnight in OptiMEM medium (Life Technologies) containing 1% heat-inactivated FBS. Compounds were incubated with cells for 90 minutes at 37°C. After addition of the FLASH detection reagent (DiscoveRx), luminescence was read using a fluorescence imaging plate reader (FLIPR Tetra; Molecular Devices, Sunnyvale, CA). The maximum signal per well was exported to generate concentration-response curves, and the EC 50 and E max values were calculated with the proprietary IC 50 -Witch software (Actelion Pharmaceuticals), using the compound intrinsic curve maximum and minimum as plateau values. Compound efficacies were compared with the maximal efficacy of iloprost (100%).
Analysis of IP Receptor Internalization by Flow Cytometry. Isogenic T-REx-HEK-hIP cells were seeded into 12-well plates, and the IP receptors were induced by cultivation in 10 ng/ml tetracycline for 24 hours to obtain IP receptor expression levels sufficiently high for flow cytometric detection. Agonist dilution series were then added in low serum medium (0.5% FBS). After incubation with agonists for the indicated times, the cells were washed with PBS and detached by rapid trypsinization and quenching with trypsin inhibitor. Then the IP receptors were stained with mouse anti-Flag antibodies (1:1000; cat. no. F-3165; Sigma-Aldrich, St. Louis, MO), diluted in PBS/2 mM EDTA and 0.5% fatty acid-free BSA (Calbiochem). As secondary antibody Alexa Fluor 488 goat anti-mouse IgG (Life Technologies) was used.

Efficacy and Desensitization Profile of Selexipag
The propidium iodide-negative cells were analyzed using the flow cytometer FACSAria IIu (BD Biosciences, Heidelberg, Germany). The median fluorescence intensity of every sample was corrected for the background fluorescence of parental cells not expressing the IP receptor and was then used to quantify the IP receptor surface expression levels.
In Vivo Studies. Male Wistar and spontaneously hypertensive rats (SHR) were obtained from Harlan Laboratories (Horst, the Netherlands). All rats were maintained under identical conditions in climate-controlled conditions (18-22°C, 40% to 60% relative humidity), with a 12-hour light/dark cycle in accordance with the guidelines of the Baselland Cantonal Veterinary Office.
SHRs were pretreated with buprenorphine (Temgesic, 0.03 mg/kg; Essex Chemie AG, Lucerne, Switzerland), and anesthesia was induced and maintained by inhalation of 2%-4% isoflurane (100% O 2 ). The abdomen was then opened with a midline abdominal laparotomy, and a blood pressure-sensing catheter was placed in the descending aorta below the renal arteries, pointing upstream. The blood-pressure transmitter (TA11PA C40; Data Sciences International, New Brighton, MN) was implanted into the peritoneal cavity under sterile conditions. The transmitter was sutured to the inside of the abdominal wall. After surgery the rats were transferred into a dedicated recovery room and were monitored for 3-4 days. Buprenorphine (0.03 mg/kg s.c. injection) was administered once daily for 2 days after surgery.
A similar methodology was used for transmitter implantation in the monocrotaline-pulmonary hypertensive rats.
The blood-pressure transmitter was placed in the abdominal cavity, and the sensing catheter was positioned in the thorax by using a trocar. After removal of the trocar, the right ventricle was punctured, and the sensing catheter inserted into the right ventricle and pushed into the pulmonary artery. The rats were monitored for 4 days after surgery with administration of buprenorphine (0.03 mg/kg s.c.) once daily. Three weeks after the implantation of telemetry devices, the rats were treated with monocrotaline (MCT; Sigma-Aldrich, St. Louis, MO) as a single s.c. injection (60 mg/kg). Four to 5 weeks after injection of MCT, the rats became pulmonary hypertensive, and we evalutated the effects of repeated oral administration of selexipag on pulmonary arterial pressure.
Telemetry units were obtained from Data Sciences International. The implanted sensor consisted of a fluid-filled catheter (0.7 mm diameter, 8 or 10 cm long; model TA11PA C40) connected to a highly stable low-conductance strain-gauge pressure transducer, which measures the absolute arterial pressure relative to vacuum, and a radiofrequency transmitter. Gas-sterilized and precalibrated implants were provided by the manufacturer. Before implantation of the transmitters, calibration was verified to be accurate within 5 mm Hg. The transmitter signals were coded and monitored by a receiver (RPC-1, Data Sciences International). The signal from the receiver was consolidated by a multiplexer (Data Exchange Matrix; Data Sciences International) and was sent to a designated personal computer (Optiplex, 960; Dell, Round Rock, TX). Arterial pressures were normalized by using input from an ambient-pressure reference (APR-1; Data Sciences International).
Systemic or pulmonary arterial pressures were collected at 5-minute intervals throughout the experimental period using the Dataquest ART Gold acquisition system (version 4.3; Data Sciences International). Blood pressure signals were sampled at 500 Hz. We calculated 5-minute intervals or hourly means of mean systemic arterial pressure (MAP) or mean pulmonary arterial pressure (MPAP). The 24-hour period before treatment was used as the control period, and each rat served as its own control by using the data from the 24 hours before treatment. The heart rate was derived from the pressure waveform.
Osmotic infusion pumps (2ML1 and 2ML2 models; ALZET, Cupertino, CA) were used to deliver selexipag and treprostinil as continuous i.v. infusions. In brief, the rats were pretreated with buprenorphine (0.03 mg/kg, s.c.). and anesthetized with 2%-4% isoflurane (100% O 2 ). The jugular vein was cannulated, and the pump was implanted under the skin midscapula. The wound was closed using medical tissue adhesive and disposable skin staples.
All in vivo results are presented as mean 6 S.E.M. Maximal effects between MAP and MPAP versus control period are expressed in mm Hg.
Determination of ACT-333679 (Active Metabolite of Selexipag) and Treprostinil Plasma Concentrations. Selexipag is the parent drug of the active metabolite ACT-333679 (Kuwano et al., 2007). After i.v. administration of selexipag or treprostinil, blood samples (250 ml) were collected in 5% EDTA (K2E EDTA, BD Microtainer Ref 365975; BD Biosciences) and centrifuged at 10,000g at 4°C. The plasma samples were put in a 96-well polymerase chain reaction plate and stored at 220°C. Plasma samples were analyzed for ACT-333679 or treprostinil concentrations using liquid chromatography coupled to tandem mass spectrometry and a deuterated internal standard for ACT-333679.
The analytic equipment consisted of a Shimadzu high-pressure liquid chromatography system (Shimadzu, Reinach, Switzerland) connected to an API5000 (AB SCIEX, Concord, ON, Canada). Data acquisition was performed with the Analyst software package (AB SCIEX). The chromatographic analysis was achieved on a Phenomenex RP Polar column (4 mm, 2.0 Â 20 mm ID) for ACT-333679 or on a Phenomenex Luna C8 column (5 mm, 2.0 Â 20 mm ID) for treprostinil, at room temperature with a flow rate of 0.6 ml/min (Phenomenex, Torrance, CA). Mobile phases consisted of 0.1% aqueous formic acid (ACT-333679) or ammonium formate 5 mM pH 9.0 (treprostinil) and acetonitrile. The mass transitions used for ACT-333679, its deuterated standard, and treprostinil were 420.4-378.3, 427.4-379.4, and 389.1-331.5, respectively, all with a scan time of 50 milliseconds. The inclusion of quality control samples with acceptance criteria of 615% in the bioanalytic runs was used to check the performance of the assays.

ACT-333679 Is an Efficacious and Potent Inhibitor of PASMC Contraction, Proliferation, and Extracellular
Matrix Production. ACT-333679 was compared in terms of potency and efficacy with the prototypic PGI 2 analog iloprost using human primary PASMC. In those cells we measured a set of PAH-relevant phenotypic parameters, including cytoskeletal contraction, cell proliferation, and extracellular matrix production. PASMC relaxation is mediated by the cAMP-dependent phosphorylation of MLCK, which leads to a decrease in kinase affinity for Ca 21 /calmodulin and reduced enzymatic activity. This results in lower myosin light chain phosphorylation and activation (Horman et al. 2008). We therefore investigated whether the IP receptor agonists ACT-333679 or iloprost increased MLCK phosphorylation in PASMC.
Both ACT-333679 and iloprost resulted in a concentrationdependent increase in MLCK phosphorylation with comparable maximal efficacies and EC 50 values of ∼3 nM and ∼0.1 nM, respectively ( Fig. 2A, B). Equal maximal efficacies in PAHrelated readouts are important and indicate comparable therapeutic potential. Different potency values, on the other hand, are not relevant if the required in vivo exposures are reached.
We next analyzed whether the observed effects on MLCK phosphorylation translated to cellular shape changes by use of label-free impedance technology (Nayler et al., 2010). Both compounds induced a concentration-dependent, rapid, and pronounced decrease in impedance (shown for ACT-333679 in Fig. 2C), indicating cytoskeletal relaxation. Minimum impedance values within the first 3 hours after stimulation were used to generate concentration-response curves (Fig. 2D).
ACT-333679 and iloprost had equal maximal efficacies and displayed respective EC 50 values of 4.3 nM (n 5 3; s g 5 1.2) and 0.12 nM (n 5 3; s g 5 1.1); that is, the potency values were very comparable to those seen in the MLCK-P assay. The parent compound of ACT-333679, selexipag, had an EC 50 value of 157 nM (n 5 3, s g 5 1.3); that is, it was 37-fold less potent than its active metabolite and had a maximal efficacy comparable to that of the other two agonists. The myosin inhibitor blebbistatin decreased PASMC impedance in a concentration-dependent manner to a comparable extent and with similar onset of action and kinetics as ACT-333679, suggesting that PASMC shape changes that are induced by IP receptor agonist stimulation reflect actomyosin relaxation (Fig. 2E).
Next we analyzed whether ACT-333679 could also counteract ET-1-induced PASMC contraction. PASMC were prestimulated with 10 nM ET-1 and after reaching a response plateau, IP receptor agonists were added (shown for ACT-333679 in Fig. 2F). ACT-333679 or iloprost induced a concentration-dependent decrease of impedance with comparable maximal efficacies for both agonists and a mean potency for ACT-333679 of IC 50 13 nM (n 5 2), and for iloprost of IC 50 2.3 nM (n 5 2). Concentration-response curves were generated from impedance data at the 190-minute time point (Fig. 2G).
To visualize this cellular shape change, we monitored PASMC in real time using a tomographic microscope. ET-1 (100 nM) addition induced strong contraction of PASMC within less than 30 minutes. Addition of ACT-333679 (1 mM) after 70 minutes reverted this contraction (Fig. 3A). In contrast, addition of vehicle instead of ACT-333679 did not revert the ET-1-induced contraction (Fig. 3B), nor did it induce any cellular shape change when added at both time points (Fig.  3C). The corresponding movies can be viewed under http:// orbit.actelion.com/selexipag/. These data suggest that IP receptor agonists effectively attenuated procontractile signaling pathways and actomyosin contraction in PASMC.
We also investigated the potential antiproliferative activity of ACT-333679 and iloprost. PDGF-BB, which has been proposed to contribute to the development of PAH, induces proliferation and extracellular matrix production of vascular smooth muscle cells. PDGF-BB decreases the expression of p27 (Kip1), an inhibitor of cyclin-dependent kinases, and increases expression of cyclin D 1 (Weber et al., 1997). A 24-hour PDGF-BB treatment of quiescent PASMC strongly decreased p27 (Kip1) expression, which was concentration-dependently and potently attenuated by cotreatment with ACT-333679 or iloprost (Fig. 4A). Also, PDGF-BB treatment of PASMC strongly induced cyclin D 1 expression, which was again concentration-dependently and potently reduced by ACT-333679 or iloprost cotreatment (Fig. 4B).
Furthermore, to quantify PDGF-BB-induced PASMC proliferation, we performed [ 3 H]-thymidine incorporation assays. The normalized data of at least 10 experiments per compound were pooled, and the average values (6 S.E.M.) are presented here. Both ACT-333679 and iloprost partially reduced PDGF-BB-induced PASMC proliferation with similar maximal efficacies (∼20%) and with IC 50 values of 4.0 nM and 0.11 nM, respectively (Fig. 4C).
Thus, the nonprostanoid selective IP receptor agonist ACT-333679 displayed efficacy in all tested PAH-related phenotypic parameters in PASMC comparable to that of the
ACT-333679 Is a Partial Agonist in Receptor-Proximal cAMP Accumulation Assays. After having analyzed two IP receptor agonists in receptor-distal, PAH-relevant phenotypic assays, we investigated the activation of IP receptor-proximal intracellular signaling pathways in PASMC. To that end, we analyzed the cAMP increase in PASMC in response to selexipag, ACT-333679, iloprost, and the additional PGI 2 analogs beraprost and treprostinil.
All IP receptor agonists induced cAMP synthesis in a concentration-dependent and saturable manner (except the parent drug selexipag, which did not reach an efficacy plateau) (Fig. 5A). Interestingly, ACT-333679 behaved as partial agonist with a maximal efficacy of 56% when compared with the three PGI 2 analogs, all of which had similar maximal efficacies of ∼100%. Intrinsic compound potencies ranged from 16 nM (iloprost) to 214 nM (ACT-333679) (Table 1). Thus, in receptor-proximal cAMP synthesis partial agonism of ACT-333679 was uncovered, a property not seen when measuring receptor-distal phenotypic parameters in PASMC. In addition, due to its receptor-proximal position in the signaling cascade, the potency of the agonists in the cAMP readout was ∼50to 100-fold lower than in receptor-distal assays.
The appearance of partial agonism depends not only on the position of a readout in the signaling cascade but also on the cellular receptor reserve-that is, the cellular receptor expression levels. We thus characterized the different IP receptor agonists in cells with high and low receptor expression levels using a tetracycline-inducible T-REx-HEK-hIP cell line.
Cell treatment with the lowest (0.1 ng/ml) and the highest (10 ng/ml) concentration of tetracycline resulted in a 60-fold difference in receptor expression as analyzed by immunoblotting (Fig. 5B). Parental cells did not show a cAMP response (Fig. 5C, left panel). In HEK-high hIP cells, all compounds behaved as full agonists (Fig. 5C middle panel, Table 1) with a potency of 0.53 nM for iloprost, 29 nM for ACT-333679, and 813 nM for selexipag. In contrast, in HEK-low hIP cells (Fig.  5C, right panel, Table 1) the nonprostanoid agonists displayed partial agonism (E max 45% for ACT-333679; selexipag did not reach saturation), whereas the PGI 2 analogs were full agonists. Furthermore, in HEK-low hIP cells all five agonists displayed potencies that were comparable to those observed in PASMC ( Fig. 5A and Table 1).
We conclude that ACT-333679 (and likely also selexipag) is a partial agonist at the recombinant and naturally expressed IP receptor, and this property is only uncovered in cell systems with low receptor expression by use of a receptor-proximal readout. Previous cAMP measurements in recombinant cells expressing the IP receptor had not detected this partial agonism, most likely due to higher IP receptor expression levels (Kuwano et al., 2007). In addition, selexipag is 28-fold less potent than the active metabolite ACT-333679 (cAMP assays in HEK-high IP), comparable to the factor observed in the PASMC impedance assays (37-fold). This shows that in both recombinant and primary cells, the metabolite ACT-333679 is much more potent than selexipag, which explains why ACT-333679 is regarded as the main contributor to pharmacologic effects in vivo.
ACT-333679 Has Low b-Arrestin Recruitment and IP Receptor Internalization Activity. Lack of receptor internalization/desensitization would be a favorable property for chronically dosed IP receptor agonists. GPCR agonists with partial agonism in second messenger measurements have previously been shown to display partial agonism in other receptor-proximal molecular events such as b-arrestin recruitment and receptor internalization (Clark et al., 1999), processes that usually render cells insensitive to further stimulation. To investigate this aspect, ACT-333679, selexipag, and the PGI 2 analogs beraprost, treprostinil, and iloprost were characterized in b-arrestin recruitment assays.
The PGI 2 analogs behaved as high efficacy agonists (iloprost and beraprost, E max 90-100%; treprostinil, E max 67%). In contrast, the nonprostanoid agonists ACT-333679 (E max 40%) and selexipag (E max 24%) displayed significantly reduced maximal b-arrestin recruitment efficacy. The intrinsic potencies of iloprost and ACT-333679 were similar, with EC 50 values of 35 nM and 51 nM, respectively. Treprostinil and beraprost showed EC 50 values of ∼200 nM. Selexipag displayed low potency with EC 50 around 800 nM (Fig. 6, Table 2). To determine whether the differences in b-arrestin recruitment efficiency caused different degrees of receptor internalization, we treated CHO-hIP cells with IP receptor agonists and analyzed the IP receptor localization by use of immunofluorescence microscopy.
All cells displayed a moderate degree of constitutive receptor internalization even in the absence of agonist treatment (vehicle, Fig. 7). However, treatment with the PGI 2 analogs induced a strong perinuclear accumulation of receptors and depletion from the cell surface. This internalization was concentration dependent. Detectable internalization was observed for iloprost starting at 100 nM and for beraprost and Efficacy and Desensitization Profile of Selexipag treprostinil at 1 mM, which is in good agreement with the efficient b-arrestin recruitment observed at these concentrations (Fig. 7).
In contrast, the nonprostanoid agonists ACT-333679 and selexipag did not induce detectable intracellular IP receptor accumulation, even at the highest tested concentration of 10 mM and did not quantitatively deplete IP receptors from the plasma membrane. These findings were confirmed by flow cytometric quantification of cell surface receptor expression in cells treated with the various agonists (Fig. 8), where the PGI 2 analogs induced concentration-and time-dependent reduction Fig. 9. Effect of repeated oral administration of selexipag on MPAP over 5 days in conscious MCT pulmonary hypertensive rats. Selexipag (10 mg/kg) was administered bidaily (morning and evening) by oral gavage over 5 days. MPAP was measured using implanted telemetry systems. Data are presented as mean 6 S.E.M., and n = 7-8. Fig. 10. Effects of i.v. administration of selexipag over 7 days (A) and selexipag (B) and treprostinil (C) over 48 hours on MAP in conscious SHR. Vehicle or test compounds were administered continuously by i.v. infusion from implanted osmotic mini-pumps. No hemodynamic recording was possible during surgical implantation of the mini-pumps, as shown by a break in the data collection (B and C). The 48-hour period before infusion of vehicle or test compounds was used as the control period for each animal. Data are presented as means 6 S.E.M.; selexipag, n = 9; treprostinil, n = 15. of surface receptor expression while ACT-333679 did not. In fact, ACT-333679 increased the IP receptor surface expression, a phenomenon known for noninternalizing GPCR ligands, mostly antagonists (Wüller et al., 2004). Thus, in summary, while displaying full efficacy on PAHrelated phenotypic parameters, the selexipag metabolite ACT-333679 has a low b-arrestin recruitment and receptor internalization activity.
Selexipag Does Not Cause Tachyphylaxis In Vivo. The effects of repeated oral administration of selexipag on MPAP were determined in the rat monocrotaline model of pulmonary hypertension (Fig. 9). The effects of bidaily oral administration of a submaximal dose of selexipag (10 mg/kg) were measured over 5 days. Selexipag decreased MPAP after each administration for ∼2 hours. Maximal decreases in MPAP were 21 6 6 mm Hg (morning administration) and 21 6 3 mm Hg (evening administration) when compared with the 1-hour control period. There was no significant difference between the effect of selexipag on MPAP after morning and evening administration throughout the duration of the experiment (P . 0.05).
Because the time interval (approximately 12 hours) between bidaily oral administration of selexipag may have been sufficient to allow for IP receptor resensitization, the effects of continuous i.v. infusion of selexipag were determined in conscious SHR by measuring MAP. The hemodynamic effect of selexipag (1 mg/kg/h) was sustained for the whole experimental period of 7 days (maximum decrease in MAP 25 6 5 mm Hg), and MAP values returned to predrug levels after cessation of treatment and no rebound effect was observed (Fig. 10A).
In the same model the hemodynamic effects of selexipag were compared against treprostinil (Fig. 10, B and C). Continuous infusion of selexipag (1 mg/kg/h) decreased MAP maximally by 21 6 8 mm Hg over the observed 48 hours with no development of tachyphylaxis. The plasma concentrations of ACT-333679 are shown in Table 3. In comparison, treprostinil (30 mg/kg/h i.v.) decreased MAP maximally by 30 6 3 mm Hg. The maximal hemodynamic effect of treprostinil was similar to that of selexipag, but MAP levels returned to predrug levels within 36 hours despite continuous infusion and continued presence of treprostinil in plasma (Table 4).

Discussion
IP receptor agonists elevate intracellular cAMP levels and inhibit contraction and proliferation of PASMC. However, agonists of G protein-coupled receptors can also trigger receptor internalization/desensitization, leading to loss in efficacy or a need for higher agonist concentrations. Indeed, PGI 2 analogs such as cicaprost and iloprost induce IP receptor internalization (Smyth et al., 2000;O'Keeffe et al., 2008). Consistent with such in vitro findings, it has been proposed that the clinical efficacy of PGI 2 and its analogs might diminish over time, a phenomenon termed tachyphylaxis, requiring that the dose be increased to maintain efficacy (Rubin et al., 1990;Archer et al., 1996;Shapiro et al., 1997;Barst et al., 2003). Time-dependent loss of vasodilator efficacy of iloprost in perfused rabbit lungs supports these clinical findings (Schermuly et al., 2007).
Analysis of the receptor activation and desensitization profile of ACT-333679 revealed a potent and fully efficacious activation of the receptor-distal events of the cAMP pathway, namely PASMC relaxation, antiproliferation, and antifibrosis. However, markedly reduced maximal efficacy in the receptorproximal events such as cAMP accumulation, b-arrestin recruitment, and IP receptor internalization was observed (Fig. 11). In contrast, the PGI 2 analogs beraprost, treprostinil, and iloprost were full agonists showing maximal efficacy on receptor-proximal assays. Consistent with the limited in vitro desensitization potential, selexipag induced sustained vasodilation in two rat disease models while the PGI 2 analog treprostinil displayed tachyphylaxis (i.e., loss of effect) upon continued administration.
Efficacy/potency differences in G protein signaling versus desensitization have previously been observed for angiotensin receptor type 1 agonists, sphingosine-1-phosphate receptor type 1 agonists, opioid receptor agonists, and G proteincoupled receptor 109A (GPR109A) agonists (Rajagopal et al., 2006;Semple et al., 2008;Walters et al., 2009;Violin et al., 2010;Schmid et al., 2013;Zhou et al., 2013;Gatfield et al., 2014;Raehal and Bohn, 2014). Several mechanisms can explain such so-called biased GPCR agonism. First, an agonist might stabilize only a subset of receptor conformations, which then leads to preferential activation of a subset of downstream pathways. Second, differential agonist resistance to intracellular degradation can lead to a bias in the apparent GPCR activation/desensitization ratio (Gatfield et al., 2014). Third, biased behavior can be displayed by partial GPCR agonists having limited efficacy in receptor-proximal readouts such as b-arrestin recruitment and internalization but full efficacy on receptor-distal phenotypic parameters.
The biased behavior of ACT-333679 is explained by the third mechanism. ACT-333679 as partial agonist displays submaximal ability to convert its target receptor into the active state even at full occupancy. This is reflected in the partial activation of receptor-proximal processes such as cAMP accumulation, b-arrestin recruitment, and receptor internalization. However, the partial activation of receptorproximal signaling does not limit efficacy in therapeutically relevant receptor-distal phenotypic readouts due to signal amplification along the signaling cascade (Fig. 11). This is exemplified by the equal maximal efficacies of ACT-333679 and iloprost in all tested receptor-distal readouts in PASMC Efficacy and Desensitization Profile of Selexipag such as anticontraction (modulation of MLCK phosphorylation, shape change) and antiremodeling (modulation of cyclin D 1 and p27 expression, [ 3 H]-thymidine and [ 3 H]-proline incorporation). While lower maximal efficacy on such parameters would translate to lower in vivo activity, a difference in agonist potency (iloprost being 10-to 30-fold more potent than ACT-333679) does not represent a major hurdle because lower potency can be compensated by higher exposure. Signal amplification along the cAMP-protein kinase A cascade has previously been reported for platelets in which slight increases in cytosolic cAMP levels by treatment with phosphodiesterase inhibitors already lead to a strong increases in protein kinase A activity and inhibition of platelet aggregation (Seiler et al., 1987). Partial GPCR agonism causing lack of desensitization/internalization has been previously described for other GPCR-agonist systems such as the b2 adrenergic receptor, the M3 muscarinic receptor, or the chemokine receptor 5 (CCR5) (January et al., 1997;Szekeres et al., 1998;Clark et al., 1999;Oppermann et al., 1999). These studies and ours suggest that partial agonism permits circumventing receptor desensitization while still retaining efficacy in receptor-distal therapeutically relevant readouts.
Based on these in vitro results, we analyzed the IP receptor activation and tachyphylaxis potential of selexipag in different rat models of hypertension. Continuous administration of selexipag caused reduction in MPAP and MAP in rat models of pulmonary or systemic hypertension, respectively, with no evidence of tachyphylaxis in either model. Even upon continuous infusion-the most stringent method to assess desensitization in vivo-selexipag reduced MAP in SHR without any signs of tachyphylaxis. In contrast, continuous i.v. infusion of the PGI 2 analog treprostinil induced tachyphylaxis within 36 hours in SHRs despite the continued presence of the drug in plasma.
Additional mechanisms not related to differential b-arrestin recruitment may contribute to the lack of tachyphylaxis seen for selexipag. For example, the high selectivity of selexipag and ACT-333679 for the IP receptor over other prostanoid receptors (Kuwano et al., 2007;Gatfield et al., 2016) might contribute to the absence of tachyphylaxis. Indeed, continuous infusion of the nonselective PGI 2 analog iloprost elicited vasodilatation of perfused rabbit lungs that decreased over time (Schermuly et al., 2007). Iloprost can activate calcium-coupled contractile EP 1 receptors (Abramovitz et al., 2000), and tachyphylaxis was partially prevented by inclusion of an EP 1 receptor antagonist in the perfusate (Schermuly et al., 2007). These data suggest that a pharmacologic interaction between IP and EP 1 receptor subtypes contributes to the loss of efficacy to iloprost in this model. It can therefore be postulated that tachyphylaxis is lower for selective IP receptor agonists, an additional mechanism potentially contributing to the sustained efficacy of selexipag in vivo.
Taken together, the nonprostanoid agonists selexipag and ACT-333679 not only differ structurally from PGI 2 analogs, but also differ in their molecular pharmacologic effects. In the present study we show that ACT-333679 displays full anticontractile and antiremodeling efficacy comparable to that of iloprost in PASMC, but a very limited capacity for b-arrestin recruitment and IP receptor internalization in contrast to all tested PGI 2 analogs. This biased efficacy/desensitization ratio of ACT-333679 might contribute to the sustained efficacy of selexipag observed in animal models and humans (Sitbon et al., 2015). Fig. 11. Scheme suggesting how the partial agonism of the selexipag metabolite ACT-333679 at the IP receptor allows circumventing receptor desensitization while still retaining efficacy in receptor-distal anti-PAH related readouts. Receptor proximal events in the IP receptor signaling cascade such as cAMP accumulation, b-arrestin recruitment, and receptor internalization are activated by ACT-333679 with limited efficacy (partial agonism compared with the fully efficacious PGI 2 analogs) while-due to signal amplification-receptor-distal events such as anticontraction, antiproliferation, and antifibrosis are activated by ACT-333679 with full efficacy (full agonism).