Functionally Selective Inhibition of the Oxytocin Receptor by Retosiban in Human Myometrial Smooth Muscle

Abstract

Novel small molecule inhibitors of the oxytocin receptor (OTR) may have distinct pharmacology and mode of action when compared with first-generation oxytocin antagonists when used for the prevention of preterm birth. The aim was to determine the mechanism of action of small molecule OTR antagonists retosiban and epelsiban compared with the currently used peptide-based compound atosiban. Human myometrial samples were obtained at cesarean section and subjected to pharmacological manipulations to establish the effect of antagonist binding to OTR on downstream signaling. Retosiban antagonism of oxytocin action in human myometrium was potent, rapid, and reversible. Inhibition of inositol 1,4,5-trisphosphate (IP₃) production followed single-site competitive binding kinetics for epelsiban, retosiban, and atosiban. Retosiban inhibited basal production of IP₃ in the absence of oxytocin. Oxytocin and atosiban but not retosiban inhibited forskolin, and calcitonin stimulated 3′,5′-cyclic adenosine 5′-mono-phosphate (cAMP) production. Inhibition of cAMP was reversed by pertussis toxin. Oxytocin and atosiban, but not retosiban and epelsiban, stimulated extracellular regulated kinase (ERK)1/2 activity in a time- and concentration-dependent manner. Oxytocin and atosiban stimulated cyclo-oxygenase 2 activity and subsequent production of prostaglandin E₂ and F_2α. Prostaglandin production was inhibited by rofecoxib, pertussin toxin, and ERK inhibitor U0126. Oxytocin but not retosiban or atosiban stimulated coupling of the OTR to Gα _q G-proteins. Oxytocin and atosiban but not retosiban stimulated coupling of the OTR to Gα _i G-proteins. Retosiban and epelsiban demonstrate distinct pharmacology when compared with atosiban in human myometrial smooth muscle. Atosiban displays agonist activity at micromolar concentrations leading to stimulation of prostaglandin production.

Every year an estimated 15 million babies are born preterm (<37 weeks of gestation) accounting for 1 in 10 deliveries (1). Preterm birth is the leading cause of infant mortality (2) and morbidity (3) in addition to causing a significant financial burden (4) and remains an important unmet clinical challenge. One approach to treat threatened preterm delivery has been the use of inhibitors of uterine contractility to achieve tocolysis.

Inhibitors licensed for tocolytic use include salbutamol, terbutaline, magnesium sulphate, ritodrine, and atosiban (5). Atosiban is a peptide antagonist of the vasopressin V_1a and oxytocin receptors (OTRs) (6) and is used in tocolytic treatment primarily for its actions on the OTR. Oxytocin is an important stimulator of the uterus at term (7), acting initially to activate phospholipase C via Gα _q/11 to stimulate production of IP₃ and the release of intracellular Ca²⁺ from the sarcoplasmic reticulum (8). In addition to immediate Ca²⁺ release in myometrial cells, oxytocin can also stimulate longer term inflammation in the myometrium and fetal membranes via nuclear factor kappa-light-chain-enhancer of activated B cells- (9), an effect mimicked by micromolar concentrations of atosiban, and blocked by the Gα _i inhibitor, pertussis toxin (10).

The structure of atosiban is based on the peptide structure of oxytocin and it is unclear whether the second generation of small molecule antagonists, such as retosiban, which demonstrate greater selectivity for the OTR (11, 12), also demonstrate agonist activity via Gα _i at higher concentrations. Furthermore, retosiban has exhibited unexpected properties in experimental studies, intimating that its pharmacology may be different to that of atosiban. For example, retosiban inhibits stretch-induced ERK signaling in myometrial explants (13). In phase 2 trials, a single dose of retosiban significantly delayed threatened preterm labor for more than 1 week, suggesting a mechanism of action which outlasts the presence of the compound in the blood (14). In the current study, we sought to determine the pharmacological mechanism of action of retosiban compared with atosiban using a combination of in vitro approaches in both human myometrium and recombinant mammalian expression systems.

Materials and Methods

Ethical approval

All procedures involving women were conducted within the guidelines of the Declaration of Helsinki and were subject to local ethics approval (REC-05/Q2802/107). Written informed consent for sample collection was obtained prior to surgery.

Subject criteria and selection.

Subjects were recruited into a single group at elective cesarean section between 38 and 40 weeks of gestation. Subjects were not in labor (NIL), as defined by an absence of observable signs of labor including regular contractions (<3 minutes apart), membrane rupture, and cervical dilatation (>2 cm) with no augmentation.

Sample collection

Myometrial biopsies were collected at cesarean section by knife biopsy from the lower uterine segment incision and were obtained prior to administration of oxytocin. Samples were briefly washed in saline and flash frozen in liquid nitrogen, or placed in modified Krebs–Henseleit (m-KHB) solution (composition (mM): NaCl, 133; KCl, 4.7; glucose, 11.1; MgSO₄, 1.2; KH₂PO₄, 1.2; CaCl, 2.5; 2-[[1,3-dihydroxy-2-(hydroxymethyl)propan-2-yl]amino]ethanesulfonic acid, 10; pH 7.4) for contraction studies.

Cell culture

Primary myometrial cell cultures were established from whole biopsies by digestion of the extracellular matrix. Primary myocytes were isolated by 2 mg/mL collagenase (Type IV, Fisher Scientific, Loughborough, UK) digestion in Dulbecco’s modified Eagle’s medium for 1 hour at 37°C and mechanical isolation through fire-polished glass pipettes. Freshly isolated myocytes were cultured in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal calf serum and penicillin (100 IU/mL) and streptomycin (100 μg/mL). Chinese Hamster Ovary cells recombinantly expressing the human OTR (CHO-hOTR) cells were obtained from GlaxoSmithKline and were routinely maintained in F-12 media supplemented with 10% fetal calf serum, penicillin (100 IU/mL), and streptomycin (100 μg/mL). All cells were cultured at 37°C in a 95%/5% air/CO₂-humidified environment. Cells were sub-cultured at 80% confluency by lifting with 0.05% trypsin, and not used beyond passage 3.

Organ bath studies

Muscle strips approximately 10 × 2 × 2 mm were mounted vertically in 10-mL organ bath chambers in m-KHB and 95% air/5% CO₂ for isometric force recordings. Force was measured with FT03C transducers (Grass Instrument Co, Quincy, MA, USA) and recorded digitally with MacLab Chart software (ADInstruments Ltd, Oxfordshire, UK). Strips were held under 20 mN tension for 90 to 120 minutes wherein rhythmical, spontaneous contractions developed. Strips that failed to contract spontaneously were excluded. For competitive antagonism, full oxytocin-mediated concentration-response curves were generated in the presence of varying concentrations of retosiban or atosiban. Each dose of oxytocin was added for 10 minutes and separated by 20 minutes of washing.

Data analysis:

Peak responses, peak frequency, and area under the curve (integral) were calculated using MacLab Chart software from the basal and dosing periods. Changes in contractions were expressed as a percentage increase over basal and plotted using GraphPad Prism (v6.0) software.

Radioligand binding

Tissue preparation:

Frozen tissue was ground to a fine powder using a pestle and mortar and homogenized in ice-cold radioligand binding buffer (composition (mM): Tris-HCl, 50; EDTA, 2.5; and MgSO₄, 5; pH 7.4 (with KOH)). Homogenates were cleared (1000 rpm, 10 minutes, 4°C), pelleted (15 000 rpm, 15 minutes, 4°C) and resuspended in radioligand binding buffer where protein was adjusted to 1 mg/mL.

Radioligand binding assay:

Assays were performed in 500-μL volumes of radioligand binding buffer containing ~100-μg membranes and [³H]-oxytocin (specific activity: 50 μCi/mmol) (Perkin Elmer, Massachusetts, USA) at concentrations ranging from 0.001 to 50 nM. Non-specific binding was determined by the inclusion of 100 nM unlabeled oxytocin. Assays were equilibrated at 30°C for 2 hours before harvest through 0.5% polyethylenimine pre-soaked Whatman GF/B filters. Recovered radiation was determined by standard liquid scintillation counting.

Data analysis:

Specific binding was determined as total binding less nonspecific binding. Data were fitted using GraphPad Prism (v6.0) and the B_max and K_D values obtained.

Measurement of cAMP

cAMP was determined using the 2-step HTRF cAMP kit (Cisbio Bioassay, Codolet, France).

Cell culture and agonist stimulation:

Cells were seeded in 96-well plates at a density 20 000 cells/well and, where applicable, treated with 100 ng/mL pertussis toxin (PTX) for 18 hours. Cells were washed in m-KHB containing 300 µM 3-isobutyl-1-methylxanthine and equilibrated in buffer at 37°C for 15 minutes. For Gα _s experiments, cells were stimulated directly with compounds as required, and for assessment of Gα _i signaling, cells were stimulated for time period indicated prior to challenge with forskolin (FSK) (10 µM, 10 minutes). Assays were terminated by addition of 25 µL of kit-supplied lysis buffer.

Detection of cAMP:

10 µl from each well was transferred in duplicate to white, low volume 384-well plates (Corning®) and detection of cAMP proceeded exactly as per the manufacturer’s instructions (CisBio) with fluorescence determined on a PHERstar FS plate reader (BMG Labtech).

Data analysis:

cAMP levels in each well were interpolated from a 4-parameter fit of known standards (Microsoft Excel) and plotted graphically using GraphPad Prism (v6.0).

Measurement of phosphorylated ERK in cells

Levels of phosphorylated ERK (pERK) were determined using the HTRF Cellul’ERK assay kit (Cisbio Bioassay, Codolet, France) as per the manufacturer’s instructions.

Cell stimulation and lysis:

Myometrial cells were seeded in 96-well plates at a density of 5000 cells/well and cultured overnight before serum starvation for 24 hours. Where appropriate, cells were pre-treated with 100 ng/mL PTX for 18 hours. Cells were washed in m-KHB and equilibrated in buffer at 37°C for 30 minutes. Cells were challenged with compounds for time points indicated before aspiration and addition of 50 µL of supplied- lysis buffer (containing phosphatase inhibitors) for 30 minutes (room temperature).

Detection of pERK:

16 µL of each well was transferred in duplicate to white, low-volume 384-well plates (Corning®) for detection of pERK as per the Cellul’ERK assay manufacturer’s instructions with fluorescence determined on a PHERstar FS plate reader.

Data analysis:

Increases in pERK were calculated as % increase over basal (vehicle, unstimulated) cells.

Measurement of prostaglandin E₂, prostaglandin F_2α, and protein kinase A in myometrial strips

Muscle strips approximately 5 × 2 × 2 mm were mounted horizontally in a DMT Flatbed muscle strip myograph system (DMT, Hinnerup, Denmark). Strips were held under 20 mN tension in m-KHB at 37°C wherein spontaneous contractions occurred. Drugs were added directly into the organ bath as per protocol requirements. Tissue was snap frozen by rapid removal from organ baths and immediate submersion into liquid nitrogen. Concurrently, aliquots of the supernatant were collected for assessment of prostaglandin production. Where necessary, tissue biopsies were treated with 100 ng/mL PTX overnight at 4°C to inhibit Gα _i signaling, 1 hour pre-treatment with 10 μM U0126 to inhibit ERK signaling or 10 μM rofecoxib to inhibit COX-2.

Preparation of cell lysates:

Frozen tissue was diced into small pieces before homogenization in 1 mL of RIPA buffer containing protease (Pierce™ tablets) and phosphatase (phosphatase inhibitor cocktail 2 (diluted 1:100), Sigma, Poole, UK) inhibitors. Homogenization was achieved using a NEXT Advanced BBY24M Bullet Blender Storm (Next Advanced, Averill Park, NY, USA) for 10 minutes at 4°C, and the NAVY ball bearing mix (refer to manufacturer’s instructions). Lysates were subject to clearance (500 rpm, 2 minutes, 4°C) before transfer to duplicate tubes for assay and protein quantification via a modified Lowry assay (DC™ Protein Assay, Bio-Rad, CA, USA). Quantitation of prostaglandin (PG) E₂ and PGF_2α in tissue supernatant was via competitive enzyme-linked immunosorbent assay exactly as per manufacturers’ instructions. The PGE₂ assay (KGE0048) was obtained from R&D Systems (Minneapolis, MN, USA), and the PGF_2α assay (516011) from Cayman Chemicals (Ann Arbor, MI, USA). Tissue homogenates were assessed for protein kinase A (PKA) using a PKA activity assay from Arbor Assays (K027-H1) and COX-2 by sandwich enzyme-linked immunosorbent assay (DYC4198-2) from R&D Systems (Minneapolis, MN, USA).

Data analysis:

In all cases, samples were extrapolated from known standards using a 4-parameter fit curve in GraphPad prism. PGE₂ and PGF_2α production, PKA activity ,and COX-2 levels were related to protein content.

[³⁵S]-guanosine 5′-O-[gamma-thio]triphosphate binding

[³⁵S]-guanosine 5′-O-[gamma-thio]triphosphate ([³⁵S]-GTPγS) is a non-hydrolyzable analog of GTP that stays bound to activated G-proteins. Specific G-protein activation can therefore be detected by virtue of the radiolabel, and co-immunoprecipitation with specific Gα antibodies. The [³⁵S]-GTPγS assay was performed according to a modification of methodologies described previously (15-18).

Membrane preparation:

Primary myometrial cultures and CHO-hOTR cells were grown to confluence in 175-cm² flasks, and lifted and pelleted via standard cell culture techniques. Pelleted cells were homogenized in the presence of hypotonic lysis buffer (10 mM EDTA, 10 mM HEPES, pH 7.4) using a Coleman handheld homogenizer. The homogenate was pre-cleared by centrifugation (500g, 5 minutes, 4°C) and the membranes collected via centrifugation (36 000g, 30 minutes, 4°C). Membranes were resuspended in freezing buffer (10 mM HEPES, 0.1 mM EDTA, pH 7.4) at 6 mg/mL protein and rapidly frozen in liquid nitrogen. Membranes were stored at –80°C until used.

Receptor activation:

Frozen membrane aliquots were diluted to 1.5 mg/mL in assay buffer (composition (mM): HEPES, 10; NaCl, 100; MgCl₂, 10; pH 7.4) and 75 µg of membrane were added to 50 µL of assay buffer containing 1 nM [³⁵S]-GTPγS (1250 Ci mmol^-1) and 1 µM GDP, with or without ligands as required, and incubated at 30°C for 2 minutes. Non-specific binding was determined by the inclusion of 10 µM unlabeled GTPγS. Incubations were terminated by 900 µL of ice-cold assay buffer and transferred to ice. Cell membranes were recovered from the reaction mixture by centrifugation (20 000g, 6 minutes, 4°C) and the supernatant removed by aspiration. Membrane pellets were solubilized by the addition of 50 µL of ice-cold solubilization buffer (composition (mM): Tris/HCl, 100; NaCl, 200; EDTA, 1; 1.25% (v/v) Igepal CA 630; pH 7.4) containing 0.2% (w/v) SDS and vortex- mixing. Once the protein was completely solubilized, an equal volume of solubilization buffer without SDS was added.

Immuno-detection of G-proteins:

The solubilized protein was pre-cleared with normal rabbit serum (1:100 dilution) and 30 µL of Protein-G Sepharose beads (protein-G Sepharose bead suspension 30% w/v in TE buffer (10 mM Tris/HCl, 10 mM EDTA, pH 8.0)) for 60 minutes at 4°C. The protein-G beads and any insoluble material were collected by centrifugation (20 000g, 6 minutes, 4°C) then 100 µL of the supernatant was transferred to a fresh tube containing G-protein antiserum (1:100 dilution). Samples were vortex-mixed and rotated overnight at 4°C. To each sample tube was added 70 µL of Protein G-sepharose bead suspension and the samples vortex-mixed and rotated for 90 minutes at 4°C. Protein G-sepharose beads were pelleted at 20 000g and supernatant removed. Beads were washed and pelleted 3 times with 500 µL of solubilization buffer (less SDS) before re-suspension in a scintillation cocktail where [³⁵S]-GTPγS was determined by standard liquid scintillation counting methods.

Data analysis:

CPM values were expressed as a % increase over basal (unstimulated).

IP-One assay

The HTFR IP-One assay was performed exactly as per the manufacturer’s instructions (CisBio). Briefly, cells were seeded at 50 000 cells per well in white, low-volume, 384-well plates (Corning®). On the day of assay, cells were washed with stimulation buffer (provided with kits) and left with 7 µL of buffer per well. Duplicate ligand plates were made containing double the desired concentrations of antagonist, agonist, or vehicle as required and cells stimulated by direct transfer of 7 µL to corresponding wells on the assay plate. Cells were stimulated for 1 hour at 37°C before reactions were terminated and fluorescence determined on a PHERstar FS plate reader (BMG Labtech).

Data analysis:

IP-1 levels in each well were interpolated from a 4-parameter fit of known standards (Microsoft Excel) and plotted graphically using GraphPad Prism (v6.0), from where EC₅₀ values were obtained. To determine antagonist potency, pA₂ values were determined from the intercept when y = 0 using the Schild equation (Log (EC₅₀ ratio –1) vs Log [antagonist]).

Measurement of IP₃ in myometrial strips

Muscle strips were mounted horizontally in a flatbed muscle strip myograph as described above, with drugs added directly into the organ bath. Tissue was snap frozen at critical points by rapid removal from baths and immediate submersion into liquid nitrogen. Critical points included the peak of a spontaneous contraction, midway between spontaneous contractions, peak of oxytocin-mediated contraction, and during inhibition of contraction in the presence of retosiban and atosiban. Frozen tissue was mechanically ground into a fine powder by pestle and mortar and homogenized in 1 M trichloroacetic acid (TCA) for 10 minutes at 4°C in a Bullet Blender Storm (Next-Advanced, New York, USA). A 400 μM sample of TCA was transferred to tubes and 50 µL of 10 mM EDTA was added. TCA was extracted by addition of 1 mL of 1:1 (v/v) dilution of tri-n-octylamine:1,1,2-trichlorotrifluoroethane and vigorous mixing. After 10 minutes, tubes were centrifuged (1000 rpm, 2 minutes, room temperature) and 400 µL of the top aqueous phase transferred to duplicate tubes where 50 µL of 50 mM NaHCO₃ was added.

Determination of

IP₃. IP₃ levels were determined using the [³H]-IP₃ Radioreceptor assay kit (PerkinElmer Life Inc, Boston, MA) as per the manufacturer’s instructions. All dilutions were considered and IP₃ was calculated in pmol/mg protein.

Protein determination.

Unless otherwise stated, protein content was assessed by the Bradford method (19).

Statistical analysis

Contractility studies.

Experiments were repeated on biopsies obtained from different women where n represents the number of biological replicates. For Figs. 1 and 2 data were analyzed by 1-way ANOVA followed by the Dunnett post hoc test comparing test compound to vehicle, and P < .05 was considered statistically significant. For Fig. 3 data were analyzed by mixed model ANOVA with antagonist as the between measures variable followed by the Dunnett post hoc test comparing test antagonist concentration to vehicle, and P < .05 was considered statistically significant. P values were graphically represented as *P < .05, **P < .01, ***P < .001. For Fig. 4 data were analyzed by paired t-test comparing observations before and after retosiban treatment, and P < .05 considered statistically significant.

IP1 assay.

Experiments were repeated on CHO-hOTR cells where n represents the number of technical replicates. For Fig. 5 data were analyzed by mixed model ANOVA with antagonist as the between measures variable followed by the Dunnett post hoc test comparing test antagonist concentration to vehicle, and P < .05 considered statistically significant. P values were graphically represented as *P < .05, **P < .01, ***P < .001. IP-1 levels were interpolated from a 4-parameter fit of known standards and plotted graphically for calculations of EC₅₀ values using GraphPad Prism (v6.0). For antagonist potency, pA₂ values were determined from the intercept when y = 0 using the Schild equation (Log (EC₅₀ ratio –1) vs. Log [antagonist]).

IP₃ assay.

Experiments were repeated on myometrial membrane preparations from different women where n represents the number of biological replicates. For Fig. 6 data were analyzed by 1-way ANOVA followed by the Dunnett post hoc test comparing test compound to vehicle, and P < .05 was considered statistically significant. Refractory period and peak contraction were analyzed separately. P values were graphically represented as *P < .05, **P < .01.

cAMP assay.

Experiments were repeated on CHO-hOTR cells where n represents the number of technical replicates. For Fig. 7 data were analyzed by 2-way ANOVA followed by the Dunnett post hoc test comparing test compound time series to vehicle, and P < .05 was considered statistically significant. P values were graphically represented as *P < .05, **P < .01, ***P < .001.

ERK assay.

Experiments were repeated on primary human myometrial cultures derived from different patient biopsies where n represents the number of biological replicates. For Fig. 8 data were analyzed by 2-way ANOVA followed by the Dunnett post hoc test comparing test compound time series to vehicle, and P < .05 was considered statistically significant. P values were graphically represented as *P < .05, **P < .01, ***P < .001.

COX-2, PKA, PGE₂, and PGF_2α assay.

Experiments were repeated on biopsies obtained from different women where n represents the number of biological replicates. For Fig. 9 data were analyzed by 2-way ANOVA followed by the Dunnett post hoc test comparing test compound time series to vehicle, and P < .05 was considered statistically significant. P values were graphically represented as *P < .05, **P < .01. For Fig. 10 data were analyzed by 1-way ANOVA with atosiban and oxytocin analyzed separately followed by the Dunnett post hoc test comparing inhibitor compound to oxytocin or atosiban alone, and P < .05 was considered statistically significant. P values were graphically represented as *P < .05, **P < .01.

G-protein coupling assay.

Experiments were repeated on membrane preparations from primary human myometrial cultures derived from different patient biopsies where n represents the number of biological replicates. For Fig. 11 data were analyzed by 1-way ANOVA followed by Dunnett post hoc test comparing test compound to vehicle, and P < .05 was considered statistically significant. P values were graphically represented as *P < .05, **P < .01.

Results

Prior to detailed studies of the OTR signaling in myometrial strips, we first characterized the basic physiological responses to oxytocin. The addition of increasing concentrations of oxytocin to human myometrial smooth-muscle strips evoked a concentration-dependent increase in the magnitude, integral, and frequency of spontaneous contractions (Figs. 1A and 1B), whereas no changes were observed by addition of vehicle control (Fig. 1B). Concentration–response analysis (Fig. 1B) revealed EC₅₀ values of peak –8.224 ± 0.46 (5.9 nM), integral –8.053 ± 0.13 (8.8 nM), and frequency –8.229 ± 0.16 (5.0 nM). In radioligand binding studies, membranes were prepared from whole muscle biopsies and saturated with [³H]-oxytocin to permit the quantification of OTR expression. The binding of [³H]-oxytocin to membranes prepared from biopsies was saturable. Specific binding, as determined by inclusion of 100 nM oxytocin represented approximately 40% of the total binding at a K_D concentration of [³H]-oxytocin. Analysis of saturation binding curves indicated a B_max value of 8337 ± 931 fmol/mg protein, and a K_D value of 24.15 ± 6.5 nM for oxytocin (Fig. 1c).

The saturation studies confirmed previous studies that the OTR reaches high concentrations in the myometrium at term (20, 21). Under such conditions it is feasible that unliganded receptor contributes to downstream signaling. Taken together with previous studies suggesting that retosiban may be an inverse agonist (defined as a ligand that can reduce constitutive receptor activity) (13), we sought to investigate whether retosiban could reduce spontaneous contractions in the absence of oxytocin. The addition of retosiban (Fig. 2A) or atosiban (Fig. 2B) did not significantly reduce the mean of magnitude, activity integral, or frequency of contractions in spontaneously contracting strips. In competition assays using oxytocin as agonist, however, the presence of increasing concentrations of retosiban evoked a right-ward shift in the oxytocin concentration–response curve for peak (Fig. 3A), integral (Fig. 3B), and frequency (Fig. 3C). Similar observations were made with atosiban (Fig. 3D-F). EC₅₀ values observed are displayed in Table 1.

Data from a phase 2 clinical trial suggest that retosiban may have a long acting mechanism of action (14). To test whether retosiban remained active at the receptor for prolonged periods we challenged tissues with oxytocin immediately after washout of a preincubation with retosiban. The actions of retosiban were rapidly reversible. Removal of retosiban by washing and immediate addition of 100 nM oxytocin resulted in contractile responses that were instant (Fig. 4A), and of similar magnitude to those preceding retosiban addition (Fig. 4B).

To investigate potential for ligand bias (defined as a ligand-specific selectivity for a downstream signaling pathway) from the OTR antagonists, we systematically investigated their mechanism of downstream signaling and potency. To confirm that any ligand bias was not simply a myometrial cell phenomenon we first tested the effect of the OTR antagonists on a Chinese Hamster Ovary cell line stably expressing the human oxytocin receptor. Oxytocin generated concentration-dependent increases in IP-1 accumulation (Figs. 5A-C), a stable metabolite of IP₃. Maximal responses were achieved at 100 μM oxytocin with an EC₅₀ value of –6.48 ± 0.17 (332 nM). The presence of increasing concentrations of retosiban (Fig. 5A), atosiban (Fig. 5B), and epelsiban an additional potent small molecule inhibitor of the OTR (22, 23) provoked a right-ward curve shift (Fig. 5C), whereby increasing concentrations of oxytocin were required to illicit the same responses. pA₂ values (the dose of antagonist needed to shift the agonist dose–response curve 2-fold to the right) were obtained using the Schild equation (Fig. 5D). Of the 3 antagonists, epelsiban was most potent with a pA₂ value of –8.537, followed by retosiban –8.045, and finally atosiban –6.159. The data indicate an increase in potency of inhibition of IP₃ generation as approximately 77-fold for retosiban and 240-fold for epelsiban compared with atosiban. Further analysis of Schild plots revealed slopes of 1.017, 1.069, and 1.042 for epelsiban, retosiban, and atosiban, respectively. pA₂ and slope values are detailed in Table 2.

To investigate IP₃ production in intact tissue, uterine strips were treated with various compounds, and snap frozen either mid-way between or at the peak of contractions and IP₃ content determined. Initial comparisons of contraction versus inter-contraction interval demonstrated IP₃ levels were essentially the same irrespective of contractions (Fig. 6A). By contrast 100 nM oxytocin markedly increased IP₃ levels (Fig. 6A), while IP₃ was significantly reduced in the presence of 1 μM retosiban (Fig. 6A). To investigate and confirm this effect further, we increased the number of observations during peak contraction whereby retosiban again decreased IP₃ compared with vehicle control whereas IP₃ levels in 10 μM atosiban-treated strips decreased but did not reach significance (Fig. 6B).

To further investigate downstream signaling and potential ligand bias at the OTR, we challenged CHO-hOTR cells with either oxytocin, atosiban, retosiban, or epelsiban and measured cAMP accumulation. None of the ligands elevated levels of cAMP (Fig. 7A), whereas addition of either FSK or calcitonin, a selective Gα _s-coupled receptor agonist, caused a robust and saturable elevation of cAMP accumulation. These data suggest that the OTR ligands do not signal through Gα _s G-proteins in these cells.

Both oxytocin and atosiban reduced calcitonin (Fig. 7b) and FSK mediated cAMP production (Fig. 7c), providing evidence of Gα _i- coupling. The potent and selective 5-HT_1B (Gα _i-coupled) agonist CP93129 (24) evoked similar changes. Importantly, neither retosiban nor epelsiban had any effect on cAMP. The oxytocin- and atosiban-mediated inhibition of cAMP were PTX sensitive demonstrating Gα _i involvement (Fig. 7D).

Stimulation of Gα _i signaling in myometrial cells has the potential to activate ERK (25). We therefore tested whether challenge of cultured myometrial cells taken at term cesarean section with either oxytocin or atosiban evoked an increase in ERK phosphorylation. Application of either oxytocin or atosiban increased ERK activity in a concentration-dependent manner (Fig. 8A) and generated EC₅₀ values of –6.93 ± 0.34 (118 nM) and –5.10 ± 0.32 (8.03 µM) for oxytocin and atosiban respectively that peaked at ~5 minutes and subsequently declined (Fig. 8B). Cells did not respond to either controls, retosiban, or epelsiban at equivalent receptor occupancy. Oxytocin- and atosiban-mediated responses were inhibited by preincubation with PTX (Fig. 8C).

Increases in ERK activity in myometrial smooth muscle are associated with increased production of prostaglandins. We observed an increase in COX-2 (Fig. 9A) expression, and PGE₂ (Fig. 9C) and PGF_2α (Fig. 9D) secretion on challenge of spontaneously contracting myometrial strips with oxytocin and atosiban, but not retosiban or epelsiban. COX-2 expression increased rapidly, but PGE₂ and PGF_2α secretion was absent until 2 hours when a 3- to 4-fold increase was observed. The magnitude of responses was similar for both oxytocin and atosiban. No increases in PKA activity were detected (except with the positive control FSK), and no reduction in basal PKA levels were observed (Fig. 9B). PGE₂ (Fig. 10A) and PGF_2α (Fig. 10B) secretion was inhibited by PTX (Gα _i inhibition), U0126 (ERK inhibition), and rofecoxib (COX-2 inhibition). COX-2 expression was inhibited by PTX and U0126 (Fig. 10C).

The downstream signaling analysis of the ligands suggested that contrary to atosiban being a neutral antagonist, it is in fact an agonist at Gα _i at micromolar concentrations. To investigate coupling of the OTR to specific Gα- subunits we utilized the [³⁵S]-GTPγS immunoprecipitation assay in membrane preparations from cultured myometrial cells taken from patients at term. Oxytocin increased [³⁵S]-GTPγS binding to Gα _q/11 (Fig. 11A). G-proteins and both oxytocin and atosiban increased binding of Gα _i subunits (Fig. 11B). Consistent with cAMP data no binding was observed following immunoprecipitation with Gα _s G-proteins (Fig. 11C). Combined with signaling analysis these data confirm that at micromolar concentrations atosiban couples OTR to Gα _i to elicit activation of ERK in human myometrial cells. Importantly, neither retosiban nor epelsiban elicited any increases in [³⁵S]-GTPγS binding, signifying no agonist activity with these compounds.

Discussion

This study demonstrates that there are significant differences in the pharmacology of 2 small molecule OTR antagonists, retosiban and epelsiban, compared with atosiban, a peptide antagonist licensed for use as a tocolytic for the treatment of preterm labor. The observations of this study, summarized in Fig. 12 explain some, but not all, of the observed physiological effects of retosiban in the published literature. In a phase 2 trial, administration of intravenous retosiban as a single dose was sufficient to inhibit spontaneous preterm birth by more than 1 week (14). Our data suggest that it is highly unlikely that the long-lasting effect of the single dose was a consequence of continued action at the OTR, such as has been previously observed for the β ₂-adrenergic receptor agonist salmeterol (26), since antagonism of oxytocin-stimulated contractions in uterine strips was rapidly reversible.

All 3 antagonists tested in this study effectively inhibit OTR coupling to Gα _q/11 and the generation of IP₃, with epelsiban being the most potent, followed by retosiban and atosiban. Since stimulation of Ca²⁺ release from the sarcoplasmic reticulum by IP₃ is the first component of the mechanism of oxytocin’s rapid action on the uterus (27) administration of all compounds should result in inhibition of the initial contractile effect. In addition to an effect on oxytocin mediated IP₃ signaling, the addition of retosiban reduced IP₃ in intact strips under basal conditions. Such a reduction is consistent with inverse agonism although it is difficult to rule out the presence of endogenous OT bound to recycled OTR in myometrial samples or endosomal signaling during altered trafficking.

Previous work on atosiban has demonstrated that at micromolar concentrations the molecule acts as an agonist stimulating Gα _i and causes inflammation in amnion cells (9, 10). In this study we demonstrated that in the myometrium atosiban at micromolar concentrations stimulates coupling of the OTR to Gα _i and subsequent phosphorylation of ERK 1/2, COX2 upregulation, and PGE₂ and PGF_2α secretion. No such stimulation was observed with the addition of retosiban. In CHO hOTR cells, addition of micromolar atosiban or nanomolar oxytocin inhibited calcitonin and FSK stimulated cAMP production, an effect that was prevented by addition of pertussis toxin, suggesting that the observed ligand bias is not cell specific.

The functional consequences of the ligand bias of atosiban in vivo are hard to interpret without accurate tissue concentrations. Tissue concentrations that reach, or exceed, micromolar levels are likely to elicit inflammation and release of prostaglandins some hours after treatment. Thus, the tocolytic effect mediated by inhibition of Gα _q/11 may initially be effective in inhibiting contractions but after some hours would paradoxically make contractions more likely via agonism of the OTR and stimulation of Gα _i.

We conclude that the small molecule oxytocin receptor antagonists retosiban and epelsiban inhibit downstream signaling in myometrial cells and CHO-hOTR cells in a manner consistent with that of a neutral antagonist. By contrast, the peptide based, mixed oxytocin/vasopressin V_1a receptor antagonist atosiban, inhibits oxytocin receptor signaling through the Gα _q/11 G-protein but acts as a partial agonist for signaling through the Gα _i pathway.

Abbreviations

Abbreviations
 
• [³⁵S]-GTPγS

  [³⁵S]-guanosine 5′-O-[gamma-thio]triphosphate

 
• AC

  adenylate cyclase

 
• ANOVA

  analysis of variance

 
• ATP

  adenosine trisphosphate

 
• COX2

  cyclo-oxygenase 2

 
• cAMP

  3′,5′-cyclic adenosine 5′-mono-phosphate

 
• DAG

  diacylglycerol

 
• ERK

  extracellular regulated kinase

 
• FSK

  forskolin

 
• IP₃

  inositol 1,4,5-trisphosphate

 
• m-KHB

  modified Krebs–Henseleit

 
• MAP

  kinase, mitogen activated kinase

 
• OTR

  oxytocin receptor

 
• PIP2

  phosphatidylinositol 4,5-bisphosphate

 
• PGE2

  prostaglandin E2

 
• PGF2α

  prostaglandin F2α

 
• PKA

  protein kinase A

 
• PKC

  protein kinase C

 
• PTX

  pertussis toxin

 
• SEM

  standard error of the mean

Acknowledgments

We thank the patients and midwives of the Biomedical Research Unit at University Hospital Coventry and Warwickshire for the collection and curation of myometrial samples.

Financial Support: This work was funded by a project grant to A.M.B, S.Q. and P.J.B. from GlaxoSmithKline

Author Contributions: A.M.B. conceived of the project, acquired the grant, analyzed data and co–wrote the manuscript. P.J.B. undertook all experimental work, analyzed data and co-wrote the manuscript. S.Q. acquired and curated patient samples and edited the manuscript. M.J.F. co-conceived the project, analyzed data and edited the manuscript.

Additional Information

Disclosure Summary: A.M.B. has undertaken scientific consultancy work for GlaxoSmithKline and Ferring pharmaceuticals for the development of oxytocin receptor antagonists for the treatment of preterm birth.

References