Oxytocic plant cyclotides as templates for peptide Gprotein-coupled receptor ligand designJohannes Koehbacha, Margaret O’Brienb, Markus Muttenthalerc, Marion Miazzoa, Muharrem Akcand, Alysha G. Elliottd,Norelle L. Dalyd,e, Peta J. Harveyd, Sarah Arrowsmithf, Sunithi Gunasekerag, Terry J. Smithb,h, Susan Wrayf,Ulf Göranssong, Philip E. Dawsonc, David J. Craikd, Michael Freissmutha, and Christian W. Grubera,1

aCenter for Physiology and Pharmacology, Medical University of Vienna, 1090 Vienna, Austria; bNational Centre for Biomedical Engineering Science and hTheSchool of Natural Sciences, National University of Ireland, Galway, Ireland; cDepartments of Chemistry and Cell Biology, The Scripps Research Institute, LaJolla, CA 92037; dInstitute for Molecular Bioscience, The University of Queensland, Brisbane, QLD 4072, Australia; eSchool of Pharmacy and Molecular Sciences,Centre for Biodiscovery and Molecular Development of Therapeutics, Queensland Tropical Health Alliance, James Cook University, Cairns, QLD 4878,Australia; fDepartment of Cellular and Molecular Physiology, Institute of Translational Medicine, University of Liverpool, Liverpool L69 3BX, United Kingdom;and gDivision of Pharmacognosy, Department of Medicinal Chemistry, Biomedical Centre, Uppsala University, 751 23 Uppsala, Sweden

Edited by Robert J. Lefkowitz, Howard Hughes Medical Institute, Duke University Medical Center, Durham, NC, and approved October 23, 2013 (received forreview June 24, 2013)

Cyclotides are plant peptides comprising a circular backbone andthree conserved disulfide bonds that confer themwith exceptionalstability. They were originally discovered in Oldenlandia affinisbased on their use in traditional African medicine to acceleratelabor. Recently, cyclotides have been identified in numerous plantspecies of the coffee, violet, cucurbit, pea, potato, and grass fam-ilies. Their unique structural topology, high stability, and toleranceto sequence variation make them promising templates for thedevelopment of peptide-based pharmaceuticals. However, themechanisms underlying their biological activities remain largelyunknown; specifically, a receptor for a native cyclotide has notbeen reported hitherto. Using bioactivity-guided fractionation ofan herbal peptide extract known to indigenous healers as “kalata-kalata,” the cyclotide kalata B7 was found to induce strong contrac-tility on human uterine smooth muscle cells. Radioligand displace-ment and second messenger-based reporter assays confirmed theoxytocin and vasopressin V1a receptors, members of the G protein-coupled receptor family, as molecular targets for this cyclotide. Fur-thermore, we show that cyclotides can serve as templates for thedesign of selective G protein-coupled receptor ligands by generatingan oxytocin-like peptide with nanomolar affinity. This nonapeptideelicited dose-dependent contractions on human myometrium.These observations provide a proof of concept for the develop-ment of cyclotide-based peptide ligands.

circular plant peptide | peptide ligand design | uterotonic |chemical pharmacology | peptide drugs

Cyclotides are head-to-tail cyclized plant peptides containingthree conserved disulfide bonds in a knotted arrangement

known as a cyclic cystine-knot motif (1). This confers them highstability (2) and presumably improves their oral bioactivity rel-ative to their linear counterparts (3). They were first discoveredin a decoction of Oldenlandia affinis DC. (Rubiaceae) leaves, anherbal remedy used in traditional African medicine duringchildbirth (4). The observed induction of labor and shorteneddelivery time were later studied on isolated rat and rabbit uteriand on human uterine strips (4, 5). The peptides responsible forthe contractility effects (5) raised interest because they survivedboiling, presumably as a result of their unique 3D structure,which was elucidated in 1995 (6). Since then, several plant spe-cies of the coffee (Rubiaceae) (7), violet (Violaceae) (8), legume(Fabaceae) (9), potato (Solanaceae) (10) and grass (Poaceae)families (11) have been identified to produce cyclotides. Cur-rently, ∼300 sequences have been reported (12), and the pre-dicted number of >50,000 cyclotides in Rubiaceae alone (7)suggests them to be one of the largest peptide classes within theplant kingdom. Their high intercysteine sequence variability andstructural plasticity (13), together with intrinsic bioactivities,

make them interesting templates for the development of novelpharmaceuticals (14).However, five decades after the discovery of cyclotides, there

still is not any information about specific molecular targets and/or mechanisms underlying their biological activities. It is knownthat cyclotides can, at higher concentrations, disrupt phospho-lipid bilayers (15, 16), because they expose hydrophobic residueson their surface. This endows them with physicochemical prop-erties allowing for insertion into membranes and pore formation(17, 18). Although no cyclotide target receptor has been identi-fied hitherto, the observed biological activities (e.g., their ute-rotonic effects) may be explained by specific receptor-mediatedmechanisms. In mammals, including humans, uterine musclecontractility can be elicited by activation of various signalingpathways. One physiological regulator of uterine contraction isthe neuropeptide oxytocin. In uterine tissue, this peptide activatesoxytocin and vasopressin V1a receptors (19–21), two membersof the G protein-coupled receptor (GPCR) family. GPCRs are

Significance

G protein-coupled receptors (GPCRs) are promising drug tar-gets: >30% of the currently marketed drugs elicit their actionsby binding to these transmembrane receptors. However, only∼10% of all GPCRs are targeted by approved drugs. Resorting toplant-derived compounds catalogued by ethnopharmacologicalanalyses may increase this repertoire. We provide a proof ofconcept by analyzing the uterotonic action of an herbalremedy used in traditional African medicine. We identified cyclicpeptides, investigated the molecular mechanisms underlyingtheir uterotonic activity, and report an oxytocic plant peptidethat modulates the human oxytocin/vasopressin receptors. Thisnaturally occurring peptide served as a template for the designof an oxytocin-like nonapeptide with enhanced receptor selec-tivity, highlighting the potential of cyclotides for the discoveryof peptide-based GPCR ligands.

Author contributions: C.W.G. designed research; J.K., M.O., M. Muttenthaler, M. Miazzo,M.A., A.G.E., N.L.D., P.J.H., S.A., S.G., and C.W.G. performed research; M. Muttenthaler,M.A., A.G.E., N.L.D., S.G., T.J.S., U.G., P.E.D., D.J.C., M.F., and C.W.G. contributed newreagents/analytic tools; J.K., M.O., M. Muttenthaler, M. Miazzo, N.L.D., P.J.H., S.A., T.J.S.,S.W., D.J.C., M.F., and C.W.G. analyzed data; and J.K., M. Muttenthaler, S.W., U.G., P.E.D.,D.J.C., M.F., and C.W.G. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.

Data deposition: The atomic coordinates have been deposited in the Protein Data Bank,www.pdb.org (PDB ID codes 2M9O, 2MGO, and 2RU2).1To whom correspondence should be addressed. E-mail: christian.w.gruber@meduniwien.ac.at.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1311183110/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1311183110 PNAS | December 24, 2013 | vol. 110 | no. 52 | 21183–21188

PHARM

ACO

LOGY

prominent drug targets, with ∼30% of all marketed drugs actingvia modulation of these receptors (21).We used a bioactivity-guided fractionation approach com-

bined with pharmacological and structural analysis to elucidatethe mechanism underlying the oxytocic activity of cyclotides andidentified a molecular target for native cyclotides. In addition, weused cyclotides as a template to explore substitutions that en-hanced receptor binding and agonistic activity. Our observationsprovide a proof of concept that (i) naturally occurring peptidelibraries cover a chemical space that intersects with the sampledGPCRs and (ii) cyclotides can serve as a template for the designof new classes of GPCR ligands, thus opening new avenues forcyclotide-based drug development.

ResultsBioactivity-Guided Fractionation of Uterotonic Plant Cyclotides. Anherbal extract that has been used for many generations by tra-ditional healers for its uterotonic properties was analyzed bybioactivity-guided fractionation. Dried aerial parts of O. affiniswere extracted by grinding, solvent partitioning, and solid C18-phase extraction of the aqueous filtrate to yield a crude cyclotideextract. The analysis by RP-HPLC and MALDI-TOF MSshowed that this extract contained a number of cyclotides iden-tified based on their mass, cysteine content, and hydrophobicity(7) (Fig. 1A and Fig. S1). Four subfractions eluting in the rangeof 18–54% acetonitrile were collected by preparative RP-HPLCand tested for their ability to induce contractions of humanuterine smooth muscle cells using a collagen gel contractilityassay (22) (Fig. S2). Compared with unstimulated cells, in-cubation with cyclotide-containing extracts showed a significantdecrease in the collagen gel area, which reflected an increasedcontraction of the smooth muscle cells. Further RP-HPLCfractionation generated 15 subfractions, of which six inducedsignificant contraction, ranging from 6.8–18.7% increased con-tractility over unstimulated cells (Fig. 1B).

Molecular and Pharmacological Characterization of the OxytocicCyclotide Kalata B7. The 15 HPLC fractions were analyzed byMS and tandem MS peptide sequencing, and 17 cyclotides wereidentified, with each of the fractions containing one to fivepeptides. Cyclotide sequences within biologically active fractionsdisplay sequence homology to human oxytocin, whereas cyclo-tides from inactive fractions lack any appreciable similarity(Table S1). In particular, loop 3 of the cyclotide kalata B7(-CYTQGC-) found in the most active fraction (Fig. 1B) and thesix-residue ring of oxytocin (CYIQNC-) have related sequences.Therefore, kalata B7 was isolated by RP-HPLC, and the purified

cyclotide (Fig. S3) was analyzed (i) for its ability to stimulatecontractions of uterine smooth muscle cells and (ii) for its af-finity to human oxytocin receptor or V1a receptor. Kalata B7displaced tritiated oxytocin or vasopressin in a dose-dependentmanner from the binding site of the oxytocin receptor or V1areceptor with a Ki of 50 μM and 12 μM, respectively (Fig. 2A andTable 1). It also provoked significant contraction of uterine cells(i.e., 8.4% increased contraction compared with unstimulatedcontrol cells) (Fig. 2C). We verified that the cyclotide kalata B7acted via the oxytocin receptor and/or V1a receptor on uteruscells by applying kalata B7 together with the receptor antagonistatosiban (23); this coapplication of both compounds resulted ina significant loss of contractility.If kalata B7 is an agonist at the oxytocin receptor and/or the V1a

receptor, it ought to trigger signaling via a Gq-dependent pathway.We verified this prediction by measuring the generation of inositol-1-phosphate (IP1) in response to the cyclotide in HEK293 cellsheterologously expressing either receptor. The analysis of the con-centration–response curve showed that kalata B7 was a partial ag-onist at both the oxytocin receptor and the V1a receptor (Fig. 2B)with an EC50 of 12 μM and 4 μM, respectively. Agonistic activitywas more pronounced at the oxytocin receptor (about 80% of theresponse elicited by oxytocin) than at the V1a receptor (about 40%of the response elicited by vasopressin).

Structural Characterization of Kalata B7. To understand the ligand–receptor interaction, the structure of kalata B7 was determinedby NMR (Fig. 3A and Table S2). This revealed a well-definedbackbone around the cyclic cystine-knot motif typical for Möbiuscyclotides, a type I β-turn between residues 9 and 12, a type IIβ-turn between residues 16 and 19, and a type VIa1 β-turn be-tween residues 22 and 25, as well as a β-hairpin between residues20 and 28. As shown in Fig. 3, loop 3 of kalata B7 (-CYTQGC-)is homologous to the six-residue ring sequence of human oxy-tocin (CYIQNC-). In particular, the tyrosine (Y15) and gluta-mine (Q17) residues of kalata B7 are in positions analogous tothose (Y2 and Q4) in oxytocin. NMR structural analysis of hu-man oxytocin confirmed the presence of a type II β-turn (24)(Fig. 3B). Therefore, loop 3 of kalata B7 and human oxytocinshare similarities in sequence and 3D structure. Furthermore,the structure of kalata B7 indicated that the side chains of Tyrand Gln in loop 3 protrude from the backbone (Fig. 3A). Hence,they might be capable of interacting with the oxytocin receptor.The crucial role of the tyrosine and glutamine residues (loop 3)of the cyclotide was confirmed by generating mutated variants(Y replaced by A, S, or F; Q was replaced by A or E). These wereall inactive or did not bind to the receptor (Fig. S4).

Cyclotides as Peptide Templates for Oxytocin and Vasopressin GPCRLigand Design. Cyclotides typically comprise 28–37 amino acids.Therefore, they are larger and more bulky than the nonapeptideligands oxytocin and vasopressin. Thus, we used the sequence ofkalata B7 as a template for the synthesis of oxytocin-like non-apeptides. Based on the sequence of loop 3 of kalata B7, fourpeptides were synthesized (Table 1). NMR analysis revealednegligible differences in structure relative to oxytocin, as de-termined by 1H chemical shifts (Fig. S5) and structural calcu-lations (Fig. 3 and Table S3). A comparison of the structuralensembles of the solution structure of oxytocin (Fig. 3B) with thenonapeptide kalata B7-oxytocin 1 (OT1) (Fig. 3C) revealeda similarly dynamic exocyclic tail and a defined region compris-ing residues 1–6 that overlay well (rmsd of 0.65 Å, Fig. 3D). Thesynthetic oxytocin-like peptides were tested for binding and re-ceptor activation (Fig. 4 and Fig. S4). [G5, T7, S9]-oxytocin(kalata B7-OT1) had improved binding affinity (Ki = 218 nM;Fig. 4A). This increased affinity was also evident, when assessingits ability to promote luciferase transcription in cells express-ing the human oxyocin receptor (EC50 IP1 formation = 145 nM,EC50 luciferase induction = 356 nM; Table 1), where it acted asa full agonist (compare with the maximum response of oxyto-cin and kalata B7-OT1 in Fig. 4B). Similar to the mutated

Fig. 1. Bioactivity-guided fractionation of O.affinis peptide extracts. (A) An-alytical RP-HPLC chromatogram of a peptide extract from O. affinis leaves aftersolvent extraction and in-batch C18 purification shows multiple peptide peaks,as determined by MALDI MS (Fig. S1). Fractions that were tested for biologicalactivity are labeled with dashed lines. Fraction OA44–45, showing the highestcontractility (see B), is highlighted in gray. mAU, milli-absorption units. (B)Collagen gel contractility data of partially purified cyclotide fractions (1.25mg·mL−1) on isolated human uterus smooth muscle cells. Data represent themean ± SEM. Statistical differences were analyzed using one-way ANOVA (**P< 0.01; ***P < 0.001).

21184 | www.pnas.org/cgi/doi/10.1073/pnas.1311183110 Koehbach et al.

cyclotide, the kB7-OT1 mutants (Y2A) and (Q4A) had lost theiraffinity and did not activate the oxytocin receptor (Table 1 andFig. S4). Interestingly, kB7-OT1 proved to be selective for theoxytocin receptor because it did not compete for binding ofradiolabeled vasopressin on any of the three human vasopressin(V1a, V1b, and V2) receptors at concentrations up to 10 μM (Fig.4C) in contrast to native oxytocin (25).

Uterostimulant Effects of O. affinis Extract and Kalata B7-OT1 onHuman Myometrium. The crude “kalata-kalata” plant extract andsynthetic kB7-OT1 ([G5, T7, S9]-oxytocin) were applied directly toorgan baths containing strips of human myometrium superfusedwith physiological saline solution. Application of 1 mg·mL−1 crudeextract resulted in stimulation of contraction amplitude by +8.1 ±3.5%, whereas the area under the curve increased by 313.5 ± 96%(mean ± SEM; n = 6) (Fig. 5A). Pretreatment with atosibansignificantly reduced but did not abolish the stimulatory effect ofthe extract (Fig. S6). Application of kB7-OT1 resulted in a dose-dependent increase in both contraction amplitude (1 nM: +6.0 ±3.2%; 10 nM: +41.4 ± 6.4%; 100 nM: +73.2 ± 5.4%) and areaunder the curve (1 nM: +2.7 ± 2.3%; 10 nM, +48.8 ± 20.3%;100 nM; +218.3 ± 143.6%; n = 3), which was inhibited by

pretreatment with atosiban (for amplitude, 1 nM: −0.8 ±2.7%; 10 nM, −13.6 ± 7%; 100 nM: −2.8 ± 5.8%; for area underthe curve, 1 nM: −14.5 ± 6.9%; 10 nM, −25.5 ± 4.8%; 100 nM:−11.1 ± 13.7%; n = 3) (Fig. 5 B and C).

DiscussionA decoction of O. affinis induces strong uterine contractionsafter either oral administration as a tea or intravaginal in-stillation (4, 5). In line with this contractile activity, we identifiedan active principle in peptide-containing extracts of O. affinis andHPLC-purified fractions containing various cyclotides based onthe following criteria: (i) The peptide mixtures and purifiedkalata B7 elicited contractions of uterine muscle cells that wereantagonized by the oxytocin receptor blocker atosiban; (ii) theydisplaced radiolabeled oxytocin from its heterologously expressedcognate receptor and vasopressin from the V1a receptor, theclosest relative of the oxytocin receptor; (iii) consistent with theoxytocin receptor being a Gq-coupled receptor, kalata B7 alsotriggered the formation of the canonical signaling downstreamcascade, resulting in the increased accumulation of IP1; and (iv)in organ bath experiments, both the O. affinis extract and thenonapeptide kB7-OT1 ([G5, T7, S9]-oxytocin) augmented

Fig. 2. Receptor pharmacology and bioactivity of isolated kalata B7. (A) Binding data were obtained by measuring the displacement of radioactive[3H]oxytocin (2 nM) or [3H]arginine-vasopressin (0.75 nM) by kalata B7 (10 nM to 100 μM) and control peptide oxytocin (OT) or vasopressin (AVP) (0.1 nM to 10μM) from 30 to 100 μg of human oxytocin receptor (solid lines) or V1a receptor (dashed lines) membranes. (B) Receptor activation was measured by recordingintracellular IP1 accumulation upon stimulation with kalata B7 (1–100 μM) or control peptide (0.1 nM to 10 μM) (same labels as in A). Data were fitted bynonlinear regression (sigmoidal, variable slope) and are shown as the mean ± SEM of two to three independent experiments. Binding data were normalizedto the percentage (%) of maximal binding; the 100% value refers to an average of 1.57 pmol of ligand bound per milligram of membrane for the oxytocinreceptor and 0.97 pmol bound per milligram of membrane for the V1a receptor. Fold induction of intracellullar IP1 accumulation above baseline was nor-malized to the number of cells. Ki and EC50 values are listed in Table 1. (C) Collagen gel contractility of pure kalata B7 (dark gray) at 10 μM could be abolishedby cotreatment with the oxytocin- and V1a receptor antagonist atosiban (AT, light gray) at 800 nM. Statistical differences were analyzed using one-wayANOVA (*P < 0.05; **P < 0.01). n.s., not significant.

Table 1. Pharmacology of kalata B7 and oxytocin-like peptides

Peptide Sequence

Binding affinity IP1 formation

Ki, M* EC50, M

OT receptor V1a receptor OT receptor V1a receptor

Kalata B7 Cyclo-GLPVCGETCTLGTCYTQGCTCSWPICKRN 5.0 ± 1.1 × 10−5 1.2 ± 0.1 × 10−5 1.2 ± 0.4 × 10−5 4.8 ± 1.0 × 10−6

[G5,T7,S9]-OT (kB7-OT1) CYIQGCTLS-NH2 2.2 ± 0.2 × 10−7 >1.0 × 10−5 1.5 ± 0.1 × 10−7 n.d.3.6 ± 0.4 × 10−7†

[T3,G5,T7,S8,S9]-OT CYTQGCTSS-NH2 >1.0 × 10−5 >1.0 × 10−5 >5.0 × 10−5 n.d.[T3,P4,G5,S7,S8,T9]-OT CYTPGCSST-NH2 >1.0 × 10−5 >1.0 × 10−5 >5.0 × 10−5 n.d.[P4,G5,S7,T9]-OT CYIPGCSLT-NH2 >1.0 × 10−5 >1.0 × 10−5 >5.0 × 10−5 n.d.kB7-OT1 [Y2A] CAIQGCTLS-NH2 >1.0 × 10−5 n.d. >5.0 × 10−5 n.d.kB7-OT1 [Q4A] CYIAGCTLS-NH2 >1.0 × 10−5 n.d. >5.0 × 10−5 n.d.OT CYIQNCPLG-NH2 1.8 ± 0.1 × 10−9 n.d. 3.4 ± 0.5 × 10−9 n.d.

4.2 ± 0.5 × 10−9†

Vasopressin CYFQNCPRG-NH2 n.d. 8.0 ± 0.2 × 10−10 n.d. 1.0 ± 0.1 × 10−9

*Binding affinity (Ki) and functional receptor activation (EC50) data are the mean ± SEM of two to four independent experiments. Ki values were calculatedusing IC50 values according to Cheng and Prusoff (44), with a Kd value of 1.5 nM for oxytocin (OT) on the oxytocin receptor and 0.6 nM for vasopressin on theV1a receptor. If no IC50 value has been determined, the given values represent the highest concentration tested. n.d., not determined.†Measurement of luciferase-coupled nuclear factor of activated T cells induction.

Koehbach et al. PNAS | December 24, 2013 | vol. 110 | no. 52 | 21185

PHARM

ACO

LOGY

contractions in human myometrium strips. Unlike the selectivenonapeptide, the extract also significantly stimulated frequencyof contractions and atosiban did not completely abolish stimula-tory effects, suggesting the presence of other uterotonic substancesin the plant extract. Taken together, these observations provideformal proof that the oxytocin receptor is a target of kalatacyclotides, particularly kalata B7 peptide, which was identifiedby MS and peptide sequencing as the major compound in thefraction (OA44–45) displaying the greatest contractility (Fig. 1and Fig. S1). Sequence analysis of kalata B7 showed high ho-mology between its loop 3 (-CYTQGC-) and human oxytocin(CYIQNCPLG) (Table S1). In humans, oxytocin acts on theoxytocin and vasopressin receptors and is one of the key playersin the induction of labor and uterine contraction (20). Oxytocinand V1a receptors are both expressed in the pregnant human uterusand are up-regulated during parturition, making the tissue moresensitive for stimulation (26). The endogenous ligand oxytocinis used clinically to induce labor and to prevent life-threateningpostpartum bleeding (20). Kalata B7 is a partial agonist onuterine smooth muscle cells and cells expressing the humanoxytocin and V1a receptors (Fig. 2B). The extracellular face of

both receptors is highly conserved, and it is therefore not sur-prising that many drugs engage both receptors (26, 27). In fact,the oxytocin receptor antagonist atosiban, which is used clinicallyto delay preterm birth, is also a potent antagonist at the V1areceptor (23). Adverse events reported after administration ofoxytocin and the original remedy kalata-kalata include a de-crease in blood pressure and cardiotoxic effects (4, 28); thesemight be related to the observed cross-activity, particularly if thepartial agonistic action of the effect of kalata B7 on V1a receptoris taken into account (Fig. 2B).Based on the observed pharmacological properties of kalata

B7, we performed a structural analysis to define candidate in-teraction residues with the oxytocin receptor. Cyclotides arethree times larger than oxytocin and presumably cannot enterdeep into the binding pocket of receptors. However, the NMRstructure showed that the side chains of the Tyr and Gln residuesin loop 3 protrude from the backbone, and hence are capable ofinteracting with the oxytocin receptor. To our knowledge, kalataB7 is the only cyclotide containing a tyrosine residue and a glu-tamine residue in this loop. Both residues are also present innative oxytocin. In fact, the tyrosine at position 2 in oxytocin hasbeen shown to be important for receptor–ligand interaction withresidues Y209 and F284 of the oxytocin receptor (29–31). Inaddition, loop 3 of kalata B7 contains a type II β-turn, which isalso important for the activity of oxytocin (24).Nonapeptide analogs of oxytocin are flexible, and therefore

can adopt several conformations that may allow for accommo-dating differences in the ligand binding pocket of their targetreceptors. In fact, of the several peptides that were designedusing the kalata B7 intercysteine loop 3 as a template, kB7-OT1([G5, T7, S9]-oxytocin) was found to be a selective agonist at theoxytocin receptor because it stimulated the receptor in the sub-micromolar range but did not bind to any of the other relatedreceptors (i.e., V1a, V1b, V2 receptor) up to concentrations of 10μM (Fig. 4). We confirmed that this agonist effectively stimulatedintact human myometrium (Fig. 5). The NMR data suggest thatkB7-OT1 and authentic oxytocin are very similar in their overallstructure (compare 1H chemical shifts in Fig. S5). A comparisonof the structural ensembles from the solution structures of oxy-tocin and kB7-OT1 reveals similarly dynamic exocyclic tails, withthe more defined regions of the structures overlaying well (Fig.3). The sequence of kB7-OT1 (CYIQGCTLS) is a combinationof the kalata B7-loop 3 (-CYTQGCTCS-) and oxytocin(CYIQNCPLG). Tyr2, Ile3, and Leu8 are known to be importantfor receptor recognition of oxytocin-like peptides (30). This is inline with our data (Table 1). A change of Asn to Gly in position 5of oxytocin does not impede the ability to bind and activate thereceptor but contributes to receptor selectivity. This feature hasbeen previously appreciated for oxytocin and V1a receptorantagonists (27). Based on our observations, we also consider thisof relevance in the future development of selective agonists for thehuman oxytocin and vasopressin receptors. There has only beenmodest progress made over the past two decades in identifyingselective agonists for the four receptor subtypes (23). Hence, itis, per se, of interest that a selective agonist for the oxytocin re-ceptor was discovered by extracting sequence information derivedfrom a cyclotide, revealing the intriguing possibility that ligandsfor other human receptors could be discovered in this way.At a more general level, our work provides a proof of concept

that naturally occurring peptides serve as templates for GPCRligand design (25, 32). Cyclotides represent a natural combina-torial peptide library, and they probe a chemical space that isdifficult to target by using small organic molecules that havebeen created by various synthetic strategies. Thus, at the veryleast, they can be anticipated to complement the existing col-lections of compounds that are used in drug discovery by high-throughput screening and related approaches. In fact, cyclotideshave recently been used as scaffolds to improve the stability ofpeptides that have interesting biological activities. This graftingintroduced peptide sequences into cyclotide loops and resulted inchimeric molecules, which bound to G protein-coupled receptors

Fig. 3. NMR solution structures of kalata B7 (A), oxytocin (B, red), andkalata B7-OT1 ([G5, T7, S9]-oxytocin) (C, blue). Superimposition of the 20lowest energy structures shows the backbone chain (N, C, and Cα atoms) andcysteine connectivity (yellow). The sequence of each peptide is shown abovethe structures. The structural ensemble of the side-chain orientations ofresidues Y15 and Q17 of kalata B7 is illustrated. (D) Representative ribbondiagrams of oxytocin (red) and kB7-OT1 (blue) overlaid across the backboneatoms of residues 1–6 (rmsd of 0.65 Å).

21186 | www.pnas.org/cgi/doi/10.1073/pnas.1311183110 Koehbach et al.

(33, 34), inactivated VEGF (35), stimulated angiogenesis (36),blocked entry of HIV via CXCR4 (37), and inhibited serine pro-teases (38). Here, we used the reverse approach (i.e., we extracted

the active segment of the cyclotide to create a selective ligand) anddocumented a plant peptide sharing similarity in sequence andactivity with oxytocin. The discovery of the active ergot ingredientsproduced by the fungus Claviceps purpurea was instrumental to thedevelopment of modern pharmacology. Accordingly, C. purpureahas been referred to as the treasure trove of pharmacology (39).Incidentally, ergot also contains (methyl)ergometrine/(methyl)er-gonovine (i.e., the first selective uterotonic compounds introducedinto clinical medicine). We are aware of the limitations of his-torical comparisons. Nevertheless, we believe that the rich diversityof cyclotides justifies that they also be considered as a potentialtreasure trove for drug discovery.

Materials and MethodsDetailed materials and methods are given in SI Materials and Methods.

Plant Extraction, RP-HPLC Fractionation, and Peptide Isolation. Aerial parts ofO. affinis DC. were extracted and purified as described previously (40),yielding a starting extract of peptides. Fractionation and isolation of cyclo-tides were carried out using RP-HPLC on a Dionex Ultimate 3000 unit(Thermo-Scientific Dionex).

MS and Peptide Identification. Analysis of peptides was performed on aMALDI-TOF/TOF 4800 Analyzer (AB Sciex). MS and tandemMS experiments werecarried out as described previously (8). Before MALDI analysis, samples weredesalted using C18 ZipTips (Millipore). Spectra were processed using DataExplorersoftware (AB Sciex), and cyclotides were characterized by manual peptidesequencing.

Cloning, Cell Culture, Transfection, and Membrane Preparation. Oxytocin re-ceptor and V1a, V1b, and V2 receptor cDNA sequences were inserted intopEGFP-N1 plasmids to yield C-terminal GFP fusion proteins. Preparation ofstably transfected HEK293 cell lines, propagation, and membrane isolationwere similar to previously described methods (41).

Radioligand Displacement Assays. Isolated membranes were incubated withradioactive agonists [3H]oxytocin (2 nM) or [3H]arginine-vasopressin (0.75 nM)and various concentrations of competing peptide. The reaction was stopped byfiltration over glass fiber filters using a cell harvester. Nonspecific binding wasdetermined in the presence of 1 μM oxytocin or vasopressin, respectively.

Functional Receptor Activation Assays. Luciferase-based reporter assays wereperformed as described previously (41). Briefly, HEK cells were transfectedwith firefly luciferase containing plasmid pGL4.30 luc2P. After transfection,cells were seeded into 96-well plates and incubated with logarithmicallyspaced concentrations of peptides. Following incubation, medium was re-moved and cells were frozen at −80 °C. Following cell lysis, luciferase activitywas measured using a Promega luciferase reagent kit. IP1 accumulationmeasurements were carried out using the Cisbio IP1 homogeneous time-

Fig. 4. Pharmacological selectivity of synthetic kalata B7-OT1 ([G5, T7, S9]-oxytocin) onoxytocin/vasopressin receptors. (A) Bindingof [3H]oxytocin (2 nM) tomembranesfromHEK293 cells (30–100 μgper assay) expressing thehumanoxytocin receptorwasmeasured in anexcess ofOTpeptide (0.1nMto1 μM)andkB7-OT1 (0.3nMto3μM).(B) Ability of the peptides (0.03 nMto 10 μMOTand3nMto 30 μMkB7-OT1) to signal throughGq and activate downstreamDNAbinding elements of GPCR activation inHEK293 cells stably transfected with the human oxytocin receptor was measured with a luciferase reporter gene assay. Data were fitted with nonlinear regression(sigmoidal, variable slope) andare shownas themean±SEMof three independent experiments. Bindingdata arenormalized to thepercentage (%)ofmaximal binding;the 100%value refers to an average of 1.57 pmol of ligandboundpermilligramofmembrane. Activation data are normalized to the number of cells and fold inductionabovebaseline.Ki and EC50 values are listed in Table 1. (C) Selectivity of kB7-OT1was tested on all four human receptors [i.e., the oxytocin receptor (OTR) and vasopressinV1a, V1b, and V2 receptors].Maximal binding (100%) refers to values of 1.57 pmol/mg for theOTR, 0.97 pmol/mg for the V1a receptor, 0.78 pmol/mg for the V1b receptor,and 0.30 pmol/mg for the V2 receptor, respectively. Statistical differences were analyzed using an unpaired t test (*P < 0.05; **P < 0.01; ***P < 0.001).

Fig. 5. Uterostimulant effects ofO. affinis extract andkalata B7-OT1 ([G5, T7, S9]-oxytocin) on human myometrium. Spontaneous contractions of term and non-laboring humanmyometrium superfusedwith physiological saline solution at 37 °C.(A) Application of 1 mg·mL−1 extract ofO. affinis followed by 0.5 nM oxytocin. (B)Dose–responses from 1 nM to 100 nM kB7-OT1. (C) Effects of kB7-OT1 in thepresenceof theoxytocin- andvasopressinV1a receptorantagonist, atosiban (1μM).

Koehbach et al. PNAS | December 24, 2013 | vol. 110 | no. 52 | 21187

PHARM

ACO

LOGY

resolved fluorescence assay. Cells were incubated with peptides for 1 h priorto fluorescence measurements on a SynergyH4 microplate reader (Biotek)according to the manufacturer’s recommendations.

Collagen Gel Contractility Assays. Human uterine myometrial smooth musclecells (hTERT-HM)were cultured, and collagen gels were prepared as describedpreviously (42). Gel images were taken using a Fluorchem 8900 imager(Alpha Innotech Corporation), and the gel area was measured using AlphaEaseFCsoftware (Alpha Innotech Corporation). Collagen contraction, correlating todecrease in gel area, was determined in quadruplicate. Data were statisti-cally analyzed using one-way ANOVA.

Cyclotide and Peptide Synthesis. Peptides were synthesized using Fmoc solid-phase peptide synthesis. After cleavage from resin, peptides were oxidized in0.1 M ammonium bicarbonate at pH 8.2 for 24 h and purified on RP-HPLC toyield >95% purity.

NMR Spectroscopy. Samples were dissolved in 90% (vol/vol) H2O/10% (vol/vol)D2O or 100% D2O and spectra recorded on a Bruker 600-MHz spectrometerat temperatures of 290K and 298K. The Protein Data Bank (www.pdb.org) ID

codes for kalata B7, human OT, and [G5, T7, S9]-OT (kB7-OT1) are 2M9O,2MGO, and 2RU2, respectively.

Organ Bath Myometrial Contractility Assays. Isometric force recordings weremade on strips of human myometrium obtained from the lower uterineincision site at caesarean section. All women gave informed written consentfor participation. The study was approved by the North West (Liverpool East)Research Ethics Committee (LREC 10/H1002/49) and by the Research andDevelopment Director of Liverpool Women’s Hospital National Health ServiceFoundation Trust, Liverpool, United Kingdom. Crude extract (1 mg·mL−1) orthe selective nonapeptide kB7-OT1 (1–100 nM) was added directly to theorgan bath. In some experiments, strips were pretreated with atosiban (1 μM).Contraction amplitude and area under the contraction curve were measuredand compared with spontaneous control activity (100%) using OriginPro 8.5software (OriginLab Corporation) as previously described (43).

ACKNOWLEDGMENTS. This work was supported by Austrian Science FundGrant P22889-B11 (to C.W.G.). M. Muttenthaler was supported by the EuropeanUnion Seventh Framework Programme (FP7/2007-2013) under Grant Agree-ment 254897. N.L.D. is an Australian Research Council Future Fellow. D.J.C. is aNational Health and Medical Research Council Professorial Fellow.

1. Craik DJ, Daly NL, Bond T, Waine C (1999) Plant cyclotides: A unique family of cyclicand knotted proteins that defines the cyclic cystine knot structural motif. J Mol Biol294(5):1327–1336.

2. Colgrave ML, Craik DJ (2004) Thermal, chemical, and enzymatic stability of the cy-clotide kalata B1: The importance of the cyclic cystine knot. Biochemistry 43(20):5965–5975.

3. Craik DJ (2006) Chemistry. Seamless proteins tie up their loose ends. Science311(5767):1563–1564.

4. Gran L (1973) On the effect of a polypeptide isolated from “Kalata-Kalata” (Old-enlandia affinis DC) on the oestrogen dominated uterus. Acta Pharmacol Toxicol(Copenh) 33(5):400–408.

5. Gran L, Sandberg F, Sletten K (2000) Oldenlandia affinis (R&S) DC. A plant containinguteroactive peptides used in African traditional medicine. J Ethnopharmacol 70(3):197–203.

6. Saether O, et al. (1995) Elucidation of the primary and three-dimensional structure ofthe uterotonic polypeptide kalata B1. Biochemistry 34(13):4147–4158.

7. Gruber CW, et al. (2008) Distribution and evolution of circular miniproteins in flow-ering plants. Plant Cell 20(9):2471–2483.

8. Hashempour H, Koehbach J, Daly NL, Ghassempour A, Gruber CW (2013) Character-izing circular peptides in mixtures: Sequence fragment assembly of cyclotides froma violet plant by MALDI-TOF/TOF mass spectrometry. Amino Acids 44(2):581–595.

9. Poth AG, et al. (2011) Discovery of cyclotides in the Fabaceae plant family providesnew insights into the cyclization, evolution, and distribution of circular proteins. ACSChem Biol 6(4):345–355.

10. Poth AG, et al. (2012) Cyclotides associate with leaf vasculature and are the productsof a novel precursor in petunia (Solanaceae). J Biol Chem 287(32):27033–27046.

11. Nguyen GK, et al. (2013) Discovery of linear cyclotides in monocot plant Panicumlaxum of Poaceae family provides new insights into evolution and distribution ofcyclotides in plants. J Biol Chem 288(5):3370–3380.

12. Kaas Q, Craik DJ (2010) Analysis and classification of circular proteins in CyBase. Bi-opolymers 94(5):584–591.

13. Clark RJ, Daly NL, Craik DJ (2006) Structural plasticity of the cyclic-cystine-knotframework: Implications for biological activity and drug design. Biochem J 394(Pt 1):85–93.

14. Craik DJ, Swedberg JE, Mylne JS, Cemazar M (2012) Cyclotides as a basis for drugdesign. Expert Opin Drug Discov 7(3):179–194.

15. Henriques ST, et al. (2012) Phosphatidylethanolamine binding is a conserved featureof cyclotide-membrane interactions. J Biol Chem 287(40):33629–33643.

16. Svangård E, et al. (2007) Mechanism of action of cytotoxic cyclotides: Cycloviolacin O2disrupts lipid membranes. J Nat Prod 70(4):643–647.

17. Barbeta BL, Marshall AT, Gillon AD, Craik DJ, Anderson MA (2008) Plant cyclotidesdisrupt epithelial cells in the midgut of lepidopteran larvae. Proc Natl Acad Sci USA105(4):1221–1225.

18. Henriques ST, Craik DJ (2012) Importance of the cell membrane on the mechanism ofaction of cyclotides. ACS Chem Biol 7(4):626–636.

19. Aguilar HN, Mitchell BF (2010) Physiological pathways and molecular mechanismsregulating uterine contractility. Hum Reprod Update 16(6):725–744.

20. Gruber CW, O’Brien M (2011) Uterotonic plants and their bioactive constituents.Planta Med 77(3):207–220.

21. Gruber CW, Muttenthaler M, Freissmuth M (2010) Ligand-based peptide design andcombinatorial peptide libraries to target G protein-coupled receptors. Curr Pharm Des16(28):3071–3088.

22. Condon J, et al. (2002) Telomerase immortalization of human myometrial cells. BiolReprod 67(2):506–514.

23. Manning M, et al. (2012) Oxytocin and vasopressin agonists and antagonists as re-search tools and potential therapeutics. J Neuroendocrinol 24(4):609–628.

24. Wood SP, et al. (1986) Crystal structure analysis of deamino-oxytocin: Conformationalflexibility and receptor binding. Science 232(4750):633–636.

25. Gruber CW, Koehbach J, Muttenthaler M (2012) Exploring bioactive peptides fromnatural sources for oxytocin and vasopressin drug discovery. Future Med Chem 4(14):1791–1798.

26. Maggi M, et al. (1990) Human myometrium during pregnancy contains and respondsto V1 vasopressin receptors as well as oxytocin receptors. J Clin Endocrinol Metab70(4):1142–1154.

27. Chan WY, Wo NC, Cheng LL, Manning M (1996) Isosteric substitution of Asn5 in an-tagonists of oxytocin and vasopressin leads to highly selective and potent oxytocinand V1a receptor antagonists: New approaches for the design of potential tocolyticsfor preterm labor. J Pharmacol Exp Ther 277(2):999–1003.

28. Holmes CL, Landry DW, Granton JT (2004) Science Review: Vasopressin and the car-diovascular system part 2—Clinical physiology. Crit Care 8(1):15–23.

29. Chini B, et al. (1996) Two aromatic residues regulate the response of the humanoxytocin receptor to the partial agonist arginine vasopressin. FEBS Lett 397(2-3):201–206.

30. Koehbach J, Stockner T, Bergmayr C, Muttenthaler M, Gruber CW (2013) Insights intothe molecular evolution of oxytocin receptor ligand binding. Biochem Soc Trans 41(1):197–204.

31. Zingg HH, Laporte SA (2003) The oxytocin receptor. Trends Endocrinol Metab 14(5):222–227.

32. Dutertre S, et al. (2008) Conopressin-T from Conus tulipa reveals an antagonist switchin vasopressin-like peptides. J Biol Chem 283(11):7100–7108.

33. Eliasen R, et al. (2012) Design, synthesis, structural and functional characterization ofnovel melanocortin agonists based on the cyclotide kalata B1. J Biol Chem 287(48):40493–40501.

34. Wong CT, et al. (2012) Orally active peptidic bradykinin B1 receptor antagonists en-gineered from a cyclotide scaffold for inflammatory pain treatment. Angew Chem IntEd Engl 51(23):5620–5624.

35. Gunasekera S, et al. (2008) Engineering stabilized vascular endothelial growth factor-A antagonists: Synthesis, structural characterization, and bioactivity of grafted ana-logues of cyclotides. J Med Chem 51(24):7697–7704.

36. Chan LY, et al. (2011) Engineering pro-angiogenic peptides using stable, disulfide-richcyclic scaffolds. Blood 118(25):6709–6717.

37. Aboye TL, et al. (2012) Design of a novel cyclotide-based CXCR4 antagonist with anti-human immunodeficiency virus (HIV)-1 activity. J Med Chem 55(23):10729–10734.

38. Thongyoo P, Bonomelli C, Leatherbarrow RJ, Tate EW (2009) Potent inhibitors ofbeta-tryptase and human leukocyte elastase based on the MCoTI-II scaffold. J MedChem 52(20):6197–6200.

39. Burgen A (2003) St Anthony’s gift. Eur Rev 11(1):27–35.40. Gründemann C, Koehbach J, Huber R, Gruber CW (2012) Do plant cyclotides have

potential as immunosuppressant peptides? J Nat Prod 75(2):167–174.41. Hicks C, et al. (2012) The nonpeptide oxytocin receptor agonist WAY 267,464: Re-

ceptor-binding profile, prosocial effects and distribution of c-Fos expression in ado-lescent rats. J Neuroendocrinol 24(7):1012–1029.

42. Attah AF, et al. (2012) Uterine contractility of plants used to facilitate childbirth inNigerian ethnomedicine. J Ethnopharmacol 143(1):377–382.

43. Luckas MJ, Taggart MJ, Wray S (1999) Intracellular calcium stores and agonist-inducedcontractions in isolated human myometrium. Am J Obstet Gynecol 181(2):468–476.

44. Cheng Y, Prusoff WH (1973) Relationship between the inhibition constant (K1) andthe concentration of inhibitor which causes 50 per cent inhibition (I50) of an enzy-matic reaction. Biochem Pharmacol 22(23):3099–3108.

21188 | www.pnas.org/cgi/doi/10.1073/pnas.1311183110 Koehbach et al.