Building bridges for highly selective, potent and stable oxytocin and vasopressin analogs

Oxytocin (OT) is an exciting potential therapeutic agent, but it is highly sensitive to modification and suffers extensive degradation at elevated temperature and in vivo. Here we report studies towards OT analogs with favorable selectivity, affinity and potency towards the oxytocin receptor (OTR), in addition to improving stability of the peptide by bridging the disulfide region with substituted dibromo-xylene analogs. We found a sensitive structure-activity relationship in which meta-cyclized analogs (dOTmeta) gave highest affinity (50 nM Ki), selectivity (34-fold), and agonist potency (34 nM EC50, 87-fold selectivity) towards OTR. Surprisingly, ortho-cyclized analogs demonstrated OTR and vasopressin V1a receptor subtype affinity (220 nM and 69 nM, respectively) and pharmacological activity (294 nM and 35 nM, respectively). V1a binding and selectivity for ortho-cyclized peptides could be improved 6-fold by substituting a neutral residue at position 8 with a basic amino acid, providing potent antagonists (14 nM IC50) that displayed no activation of the OTR. Furthermore, xylene-bridged analogs demonstrated increased stability compared to OT at elevated temperature, demonstrating promising therapeutic potential for these analogs which warrants further study. 2018 The Authors. Published by Elsevier Ltd. This is an openaccess article under the CCBY license (http:// creativecommons.org/licenses/by/4.0/).


Introduction
Oxytocin (OT), a nonapeptide released from the pituitary gland, is a promising therapeutic agent that is heavily involved with lactation and uterine contraction in the peripheral system and is further linked with complex neurological disorders in the central nervous system. For example, OT is the World Health Organization's recommended drug to prevent postpartum hemorrhaging, which is one of the most common causes of maternal morbidity. 1,2 However, OT suffers from limited stability in aqueous solution that is especially problematic in subtropical climates where the majority of maternal deaths occur. [3][4][5][6] Further, OT is highly susceptible to metabolic degradation, having an in vivo half-life of 3 min, causing substantial loss of activity. 7 Common approaches to overcome peptide degradation, including use of unnatural and D-amino acids, terminal capping and chemical modification or mutation of the proteolytic recognition sites, are often unsuitable for OT since its biological activity is highly sensitive to structural change. This is because OT shares a molecular structure closely related to vasopressin (also known as arginine vasopressin, AVP), making the development of a highly specific and stable oxytocin receptor (OTR) ligand challenging. Both OT and AVP contain a disulfide bridge between residues 1 and 6, resulting in a structure containing a cyclic core comprising six amino acids with a flexible three-residue amidated tail (Fig. 1, Table 1). The peptides differ by the amino acids at position 3, and at position 8 whereby OT-related peptides contain a neutral residue, while AVP peptides bear a basic amino acid. This subtle difference in polarity at position 8 is thought to confer the molecules' interaction with its receptor that, in turn, is related to its distinctive function. 8,9 However, due to their common structure, OT and AVP peptides bind to and act on multiple members of the Gprotein coupled receptor family to exert their pharmacological effects, including OTR, and AVP receptor subtypes V 1a , V 1b and V 2 . 10 For OT, Tyr2 and Asn5 are fundamental for activity in the uterus while residues Leu8, Pro7, Gln4 and Ile3 are key for receptor binding. 11 Specifically, to improve the stability of OT two main strategies have been used (i) N-terminal deamination, 12 17 and diselenide bridges. 18 Structure activity relationships (SARs) have identified that introduction of one alternative atom on the disulfide bond (such as selenium) causes a subtle change in the ring region of OT that is tolerated for OTR binding and efficacy, 15,19 while reductions in the ring size of OT can abolish OTR activity. However, the majority of studies omit determination of AVP receptor subtype activity, leaving a question over alterations in receptor selectivity induced by these chages. 19 Alternatively, carbetocin, an OT mimetic with an improved pharmacokinetic profile and prolonged uterotonic activity, 14 shows reduced selectivity and affinity towards the OTR (10-fold lower than OT), and is only a OTR partial agonist (Fig. 1). 20,21 In addition, recent reports have emerged that suggests carbetocin is also a V 1a and V 1b antagonist, complicating its pharmacological profile. 22 In a complementary and relatively less studied approach to disulfide engineering, disulfide bridging represents an attractive alternative that circumvents the lengthy and complex synthesis of unnatural cyclized peptides through engineering strategies. For example, Collins et al. recently demonstrated that a maleimidefunctionalized polymer could successfully bridge the disulfide region of OT, resulting in a conjugate with increased thermal stability. 23 In the present study, we investigated the biological activity and stability of OT analogs bridged by dibromo-xylene molecules that offer an irreversible covalent modification. Previously, dibromoxylenes have been used to assist peptide cyclisation, 24 and increase proteolytic stability or helicity of peptides, 25,26 with further applications in the construction of protein mimics. 27 Furthermore, a range of substituted analogs are available commercially, enabling bridging distance structure-activity relationships (SAR) to be studied. We produced a library of cyclized OT analogs based on N-terminally deaminated OT (dOT), which has been reported to have increased stability compared to OT. 7 In addition, omission of a Nterminal amine on dOT prevented any complications from competing side reactions during cyclization. The peptide library was screened for binding affinity, biological activity and stability at elevated temperature using in vitro assays, revealing promising potency, selectivity and stability profiles driven by xylene-bridging in OT analogs.

Synthesis of OT and peptide analogs
Native OT, deaminated OT (dOT) and dOT(L8R) were successfully synthesized via automated fluorenylmethoxycarbonyl (Fmoc) solid phase peptide synthesis (SPPS), and coupling with 2-(1H-benzotriazol-1-yl)-1,1,3,3-tetramethyluronium hexafluorophosphate (HBTU). In the case of dOT and dOT(L8R), a deaminated protected cysteine was made by reacting triphenylmethylchloride in dicholomethane with 3-mercaptopropionic acid 28 , then coupled in the final position. Peptides were cleaved from the resin using trifluoroacetic acid (TFA) in the presence of 2.5% DTT to prevent oxidation of cysteine or deaminated cysteine residues. Crude peptides were purified by reverse phase (RP) liquid chromatography mass spectrometry (LC-MS) using either water/methanol or water/acetonitrile (MeCN). When required, cyclization of native disulfide bonds was achieved by stirring the peptide in ammonium bicarbonate buffer in the presence of oxygen for up to three days. Both purified peptides were lyophilized, and characterized by RP LC-MS (Supplementary Fig. 1.1), which is presented in Table 2.

Disulfide bridging of dOT with xylene analogs
Cyclization of dOT analogs using various isomers of dibromoxylene was first attempted in a mixture of ammonium bicarbonate and acetonitrile, at room temperature and at a final concentration of 1 mM peptide. While full conversion was seen after 10 min, a small proportion of polymerized product for both meta-(11%) and para-(21%) dibromoxylene-cyclized dOT was detected. Pleasingly, this side product was removed when the reaction was performed at higher dilution (0.5 mM) and lower dibromoxylene molar equivalents (1.1 equivalents vs. 3 equivalents), affording all peptides efficiently and in good overall yield ( Table 2, Fig. 2). Xylene-bridged analogs showed reduced solubility in water; addition of 15% v/v DMSO solved this issue.

Binding affinity of peptides against OTR and AVP receptor subtypes
The OTR has a substantial reliance on cholesterol for proper functioning that has, in turn, presented severe challenges to attempts to generate a crystal structure of the OT-OTR complex. 29 Further, common problems facing solubilized OTR include reductions in affinity and loss of characteristic binding properties towards ligands. Therefore, SAR for OT analogs is best determined using in vitro receptor binding studies, with conclusive studies also testing against AVP receptor subtypes to address selectivity. Inhibitory constant (K i ) values for human receptors are presented in Table 3. Displacement of either [ 3 H]OT or [ 3 H]AVP by dOT analogs was measured over a concentration range of 0.95-30000 nM, while non-specific binding of peptide analogs was defined using the appropriate cold endogenous peptide. 30 Receptor specificity was calculated by comparing the affinity of ligands to members of the OT and the AVP receptor family member that demonstrated the highest binding capacity.
While all analogs screened suffered a decrease in binding affinity toward the OTR compared to native OT, selectivity over the AVP receptor subtypes was improved for dOT meta and maintained for the dOT para derivative (Table 3, Fig. 3a). Further, meta-cyclization demonstrated the highest affinity for OTR binding among the analogs tested. Interestingly, dOT ortho showed high affinity and preferential binding towards V 1a receptor (Fig. 3b). The introduction of a  Table 1. Table 1 Sequence residues for OT and related peptides.
basic residue at position 8 further enhanced the affinity and selectivity profile for analog dOT(L8R) ortho towards V 1a receptor compared to both dOT ortho and AVP.

Biological activity of peptides by calcium flux assay
OT/AVP peptides exert their biological function through binding to the relevant receptor, causing a measurable increase in intracellular Ca 2+ in a fluorescence assay. 29,30 Half-maximal effective con-centration (EC 50 ) values or functional antagonism (half-maximal inhibitory concentration, IC 50 ) for the peptide series is reported in Table 4. In addition to receptor binding, dOT meta and dOT para showed excellent agonist potency (EC 50 ) and selectivity towards OTR. In agreement with affinity measurements, dOT meta was the most potent and selective OTR agonist, however maximal activation was 75 ± 1%, suggesting that dOT meta is a partial agonist (Fig. 4a), but nevertheless more effective than carbetocin (45 ± 6%). 22 Surprisingly, dOT ortho showed agonist activity towards OTR    and antagonist activity towards V 1a receptor. This phenomenon has been shown recently for carbetocin, which displays agonistic behavior towards OTR and antagonism towards the V 1a and V 1b receptors. 22 Alternatively, a complete switch in receptor selectivity was demonstrated for analog dOT(L8R) ortho that displayed potent antagonist activity and selectivity towards the V 1a receptor (Fig.  4b), showing no activation of OTR. We demonstrate here that dOT meta is a selective human OTR agonist with no antagonist activity towards the OTR detected, while dOT(L8R) ortho was identified as a selective V 1a receptor antagonist. Whilst these data demonstrate that high affinity and selective binders can be achieved through disulfide bridging, modification to the OT backbone significantly influences affinity, selectivity and potency. The sensitivity of OT to alterations in ring size emphasizes the importance of screening multiple receptor subtypes when exploring OT ring modifications due to the complex pharmacological profile of this hormone. We hypothesize that modification of the disulfide bond distorts the cyclic region of these analogs, which influences their interaction with OT/AVP receptor subtypes.
Pharmacology profile of OT, AVP and xylene-bridged analogs to human OT/AVP receptor subtypes. Values are the mean value of three different experiments, each performed in duplicate.

Thermal stability
The formation of a non-reversible covalent linkage in place of a disulfide bond is expected to improve the stability of OT analogs since previous research suggests that reactions involving this region plays a central role for OT degradation, being prone to oxidation, dimerization, and b-elimination. 6 Peptide stability was   investigated through an accelerated stability test in which dOT meta was compared to OT, as a representative analog of the bridged peptides. Peptides were dissolved in water to a concentration of 0.5 mg mL À1 and heated to 50°C in a heat block for 10 days. After this time aliquots were removed and analysed by RP-LCMS. The amount of remaining peptide was calculated as a percentage of the relevant peak remaining before heat exposure. In this study, degradation was prevented by conjugation of dibromo-xylene (Fig. 5).

Conclusions
We report rapid and effective synthesis of xylene-bridged analogs of dOT and determined the impact on biological activity and stability following modification. This flexible approach facilitated access to a variety of isomeric xylene bridges, which led to diverse modulation of receptor selectivity, potency and affinity, presenting an alternative pharmacological profile and therapeutic potential to OT. A key finding from this work was that ring size and geometry influenced OTR binding, with meta-bridged analogs demonstrating the most favorable binding and pharmacology profile towards OTR. dOT meta demonstrated an improved selectivity profile for both binding and activation towards the OTR and showed minimal V 1a antagonist activity, a more selective pharmacological than the widely used drug carbetocin. 22,31 It was surprising that dOT ortho displayed preferential binding affinity and selectivity towards V 1a , which can function as both a OTR agonist and V 1a antagonist, which could be further enhanced for dOT(L8R) ortho , incorporating a basic residue at position 8, a mutation important for AVP receptor engagement. Notably, dOT(L8R) ortho displayed improved binding selectivity towards V 1a over AVP, with substantial antagonist activity only towards this receptor subtype.
OT binds and activates OTR, V 1a and V 1b receptor subtypes, which has contributed to limited or debated efficacy in clinical practice. 32,33 The molecules reported here have a defined binding and pharmacology profile, making them potentially useful tools for probing the pharmacological potential of OTR. Furthermore, xylene-bridging proved useful in enhancing thermal stability, demonstrating minimal degradation compared to native OT at elevated temperatures. Consequently, the xylene bridged analogs represent an interesting modification to OT with increased stability, worthy of further investigation.

Equipment
Peptides were purified and analysed on a Waters LC-MS system consisting of i) Waters 2767 autosampler for samples injection and collection; ii) Waters 515 HPLC pump to deliver the mobile phase to the source; iii) Waters 3100 mass spectrometer with ESI; and, iv) Waters 2998 Photodiode Array (detection at 200-600 nm)), equipped with XBridge C 18 reverse-phase columns with dimensions 4.6 mm Â 100 mm for analytical and 19 mm Â 100 mm for preparative runs. Solvents were degassed with helium and supplemented with 0.1% formic acid prior to use.

Solid phase peptide synthesis
Peptides were synthesised using automated solid-phase peptide synthesis with Rink amide Tentagel resin on a ResPep SL apparatus (Intavis) using the supplied Multipep software. Synthesis was carried out in peptide synthesis grade DMF. Resin (20 lmol/well) was swelled in DMF for 30 min before synthesis. Subsequent steps were conducted automatically. N-a-amino Fmoc groups were deprotected using 20% (v/v) piperidine in DMF (400 lL, 2 Â 5 min). The Fmoc-protected amino acid (100 lmol, 5.0 eq.; 200 lL of 0.5 M stock solution in NMP) for coupling was pre-activated with HBTU (95 lmol, 4.75 eq.; 190 lL of 0.5 M stock solution in NMP) and NMM (200 lmol, 10 eq.; 50 lL of 4 M stock solution in NMP). Coupling was allowed to take place over 30-55 min and amino acids were 'double-coupled' (i.e. the coupling step was repeated). To prevent deletion sequences, any unreacted N-terminal amines were acetylated with 'capping mixture' (400 lL of 5% (v/v) Ac 2 O in DMF) for 10 min. The resin was washed with DMF (3 Â 1 mL) between the deprotection, coupling and acylation steps. The typical cycle was repeated until the final (N-terminal) amino acid coupling, when N-a-Fmoc deprotection at the final residue undertaken under usual conditions. Following synthesis, peptides were washed several times (3 Â 1 mL DMF, 3 Â 1 mL DCM, 3 Â 1 mL MeOH, 3 Â 1 mL Et 2 O) and dried overnight in a desiccator. For the deprotection and cleavage of all peptides, a mixture of 1.5 mL TFA:H 2 O: DTT:TIS (94:2.5:2.5:1) was added to resin bound peptide for 3 h. Crude peptides were precipitated using ice cold TBME, centrifuged at 4000 rpm for 15 min at 4°C, and the supernatant discarded. The remaining peptides were washed with a fresh aliquot of TBME and the process repeated. Precipitate was dried in a desiccator over silica gel to yield off white solids which were dissolved in a H 2 O: MeOH mixture for purification by RP LC-MS. Following purification, fractions containing pure peptide were combined and concentrated in the Genevac. Subsequently, pure peptides were redissolved in water and freeze-dried overnight.

S-S disulfide bridging
Cyclization between internal cysteine residues was mediated in ammonium bicarbonate buffer (0.1 M, pH 8) at a final peptide concentration of 0.1 mg/mL. The solution was stirred for up to 3 days in the presence of oxygen at room temperature, after which time the solution was concentrated then lyophilized to yield pure peptide.

Xylene bridging of cysteine residues
A solution of dibromo-xylene (1.1 eq.; 0.55 mL from a 10 mM stock in MeCN) was added dropwise to a solution of dOT peptide in a mixture of ammonium bicarbonate (0.02 M, pH 7.8) and MeCN (3:1 respectively), affording an overall peptide concentration of 0.5 mg/ml. The resulting solution was agitated at room temperature for 30 min, and peptide products were purified and stored as a solution of water with 15% v/v DMSO.

Stability studies
Peptides were dissolved in water to a concentration of 0.5 mg mL À1 and incubated at 50°C. Aliquots of the samples were removed after 10 days and analysed after direct injection (2 Â 20 lL injection) by LC-MS.

Membrane preparation
Human receptors were cloned by RT-PCR from total human liver RNA (V 1a ), kidney RNA (V 2 ), or mammary gland RNA (OTR). Cell membranes were prepared from HEK293 cells transiently transfected with expression vector coding for human V 1a , human V 2 , or mouse V 1a . For human OTR membrane preparation, a stable HEK clone expressing the receptor was selected. The transient or stable cells were grown in 20 L fermenters. For each receptor, 50 g of cell pellet was resuspended in 30 mL of ice cold lysis buffer (50 mM HEPES, 1 mM EDTA, and 10 mM MgCl 2 adjusted to pH 7.4, with the addition of complete cocktail of protease inhibitor (Roche Diagnostics) and homogenised with Polytron for 1 min. The preparation was centrifuged 20 min at 500 g at 4°C, the pellet discarded, and the supernatant centrifuged for 1 h at 43,000 g at 4°C (19,000 rpm). The pellet was resuspended in lysis buffer and sucrose (10%). The protein concentration was determined by the Bradford method and aliquots stored at À80°C until use.

Binding affinity measurement
For vasopressin receptor binding studies, 60 mg of yttrium silicate SPA beads (Amersham) were mixed with an aliquot of membrane in binding buffer (50 mM Tris, 120 mM NaCl, 5 mM KCl, 2 mM CaCl 2 , and 10 mM MgCl 2 ) for 15 min with mixing. 50 lL of bead/membrane mixture was then added to each well of a 96 well plate, followed by 50 lL of 4 nM 3 H-AVP (American Radiolabeled Chemicals). For total binding measurements, 100 lL of binding buffer was added to the respective wells; for nonspecific binding, 100 lL of 8.4 mM cold vasopressin or cold oxytocin for hOTR was added; and for compound testing, 100 lL of a serial dilution of each compound in 2% DMSO was added. The plate was incubated for 1 h at room temperature, centrifuged 1 min at 1000 g, and counted on a Packard Top-Count.
Binding to human OTR was measured by filtration binding using 1 nM 3 H-OT final concentration in 50 mM Tris, 5 mM MgCl 2 , and 0.1% BSA (pH 7.4) buffer containing membranes. After compound addition as described above and 1 h of incubation at room temperature, the binding was terminated by rapid filtration under vacuum through GF/C filters, presoaked for 5 min with assay buffer, and washed 5 times with ice-cold assay buffer before counting. Nonspecific binding counts were subtracted from each well and data normalized to the maximum specific binding set at 100%. K i values were calculated using the Cheng-Prussoff equation. Saturation binding experiments performed for each assay indicated that a single homogeneous population of binding sites was being labeled. For receptor binding affinity (K i ) determination, compounds were tested at least 3 times in duplicate. 4.7.3. Stable cell culture and calcium flux assay using fluorescent imaging CHO cells were stably transfected with expression plasmids encoding one of the human receptor of interest and grown in F-12 K, containing 10% fetal bovine serum, 1% penicillin-streptomycin, 1% L-glutamate, and 200 lg/mL geneticin at 37°C in a 10% CO 2 incubator at 95% humidity. Cells were plated for 24 h at 50,000 cells/well in clear bottomed 96 well plates and were dye loaded for 60 min with 2 lM Fluo-4-AM in assay buffer. After cell washing, the plate was loaded on a fluorometric imaging plate reader (FLIPR), compound dilution series added to the cells, and the calcium signal recorded for 5 min in order to detect agonist activity. After 20 min of incubation with compound, a concentration natural agonist (oxytocin or AVP depending on the receptor) giving 80% of the maximum signal was added to the plate and the calcium signal recorded for 5 min in order to detect antagonist activity of the test compound.
The calcium signal reduction due to the antagonist activity of the compounds was fitted to a single site competition equation with variable slope and formula Y = Bottom + (Top-Bottom)/(1 + 10^((LogIC 50 -X)⁄HillSlope)), where Y is the% normalized fluorescence, Bottom is the minimum Y, Top is the maximum Y, IC 50 is the concentration inhibiting 50% of the agonist induced fluorescence, X is the logarithm of the concentration of the competing compound, and Hillslope the Hill coefficient. All compounds were tested at least 3 times in duplicate