Based on previous methods for cyclotide extraction and identification (Hashempour et al., 2013; Koehbach et al., 2013a; Hellinger et al., 2015a), we isolated and characterized cyclotides from ipecac root (Figure 1A) powder (C. ipecacuanha) with the aims to expand our knowledge of cyclotide sequence diversity and to investigate their modulating effects on signaling of class B GPCRs. The initial aqueous root extract of ipecac was prepared by maceration; this extract was then pre-purified by C₁₈ solid-phase extraction to remove polar plant constituents. This resulting ‘ipecac extract’ was analyzed by RP-HPLC to confirm the presence of absorbance signals across a linear acetonitrile gradient (Figure 1B). Due to the hydrophobic surface properties of cyclotides, they elute late in reversed-phase chromatography. Hence we manually collected the early- (hydrophilic) and late-eluting (hydrophobic) fractions to analyze the presence of cyclotides in the ipecac extract via MALDI-TOF MS. As expected, mass signals characteristic for the presence of cyclotides were identified within the molecular weight range of 2500–4000 Da (Figure 1C), but were absent on the more hydrophilic fraction of the extract (Figure 1D). Given our interest in the discovery of novel GPCR-modulating compounds, we measured effects of the ipecac extract on CRF₁ receptor signaling.

The ipecac extract was tested for its modulating effect on CRF₁R signaling using a cell-based luciferase reporter assay. HEK293 cells transiently transfected with an C-terminally GFP-tagged human CRF₁R and a luciferase-coupled cAMP response element (CRE) were stimulated with varying concentrations of ipecac extract (0.01–50 μg mL^-1) in the presence and absence of 50 nM endogenous corticotropin-releasing factor (CRF) ligand. Concentration-dependent inhibition of receptor signaling was determined by measuring luciferase induction and accounting for the fluorescence of receptor-positive cells. The amount of luminescence of the control samples incubated with hormone alone was defined as 100% efficacy (Figure 1E). The inhibitory potency of the corresponding cyclotide-containing ipecac extract was measured and normalized relative to the efficacy of the control. The ipecac extract exhibited CRF-stimulated CRF₁R inhibition in a concentration-dependent manner, whereas the extract only did not induce any effect (Figure 1E). The inhibitory potency of the ipecac extract to antagonize CRF₁R signaling was determined as half-maximal concentration (IC₅₀) as 2.0 μg mL^-1 (Figure 1F). Assuming that the ipecac extract contained cyclotides, it was reasonable to further determine the typical structural properties of these peptides, i.e., the presence of six cysteines and a circular peptide backbone.

Given the antagonism of CRF₁R-mediated luciferase activity by the ipecac extract, we sought to identify the cyclotides present in the extract. The ipecac extract was subject to bio-chemical derivatization by thiol-reduction, carbamidomethylation and endopeptidase treatment, using dithiothreitol, iodoacetamide, and endoproteinase (endo-)GluC, respectively. Cyclotides generally, and particularly caripe cyclotides, contain a single conserved glutamic acid residue (Koehbach et al., 2013a). Hence, digestion with endo-GluC is a valuable tool for analyzing cyclotides. It was clear that the mass signals (e.g., 3254.3, 3238.4, and 3302.3 Da) in the unmodified extract (Figure 2A) shifted by 348 ± 0.1 Da (i.e., 3602.4, 3586.4, and 3650.3 Da) after reduction and alkylation (Figure 2B), characteristic of peptides containing six thiol-oxidized cysteine residues. In addition, the masses of these peptide signals increased by another 18 Da (i.e., 3620.4, 3604.4, and 3668.3 Da) after enzymatic digest with endo-GluC (Figure 2C), corresponding to the addition of a H₂O molecule, typical for the enzymatic ring-opening of a backbone cyclized peptide.

After confirming the presence of cyclotides in the ipecac extract, we aimed to isolate the cyclotides responsible for CRF₁R antagonism using a bioassay-guided fraction approach. Despite the presence of cyclotide in the ipecac extract, the HPLC chromatogram of the ipecac extract (Figure 3A) indicated the presence of early-eluting, hydrophilic non-cyclotide compounds. To remove these unwanted impurities and for improvement of preparative cyclotide fractionation yield, the ipecac extract was pre-fractionated by solid-phase extraction as explained in the section “Materials and Methods” (Figures 3A,B). The resulting cyclotide-enriched ipecac fraction was subject to fractionation by preparative HPLC, and manual collection of seven cyclotide fractions as indicated in alphabetical order (A–G) in the chromatograms (Figure 3B). All fractions were analyzed by analytical HPLC and MALDI-TOF MS. Fractions A and B only contained trace amounts of cyclotides, whereas fractions E, F, and G each contained individual cyclotides with high purity (Supplementary Figure 1). Fractions C and D were determined to contain at least two cyclotides (Figures 4B,C and Supplementary Figure 1). All fractions were examined with the luciferase activity assay, as described earlier, and tested for agonism/antagonism at the CRF₁R. We tested two concentrations per fraction (10 and 50 μg mL^-1) in the absence and presence of CRF ligand (Figure 3C). None of the fractions mediated any changes in luciferase signal when applied to the CRF₁R cells alone. On the other hand all fractions A–G of the cyclotide-enriched ipecac fraction were capable of inhibiting the CRF-stimulated (50 nM) luciferase induction, suggesting that the cyclotides antagonize, but do not activate the CRF₁R. Analyzing the percentage of inhibition, fraction F appeared to be the most effective antagonist, which inhibited CRF-induced luciferase signal by 37.6 and 98.2% at 10 and 50 μg mL^-1, respectively. These results suggested that the ipecac extract, and the cyclotide-enriched fraction, respectively, contain cyclotides that inhibit the CRF₁R with varying inhibitory potential. It was therefore of major importance to characterize these cyclotides at a molecular level, by further purification and amino acid sequencing.

Analytical RP-HPLC and MALDI-TOF MS analyses confirmed the purity of all fractions except C and D (Supplementary Figure 1). MALDI-TOF MS spectra of the latter two fractions indicated the presence of multiple mass signals corresponding to at least two cyclotides in each fraction. Since preparative scale RP-HPLC system was not able to discriminate and resolve these co-eluting peptides (Figure 4A), we utilized a semi-preparative method to isolate these Carapichea ipecacuanha (caripe) cyclotides to purity. Fraction C eluted between 78–83 min and was collected as two sub-fractions C1 (78–79 min) and C2 (80–82 min) (Figure 4B). MALDI-TOF MS spectra and analytical HPLC analysis of each fraction confirmed their high purity, and indicated the presence of a single, distinct cyclotide in each fraction, caripe 10 and caripe 11, respectively (Figure 4B and Supplementary Figure 1). The semi-preparative HPLC chromatogram of fraction D contained a broad peak eluting between 83 and 87 min, which was collected as two sub-fractions D1 (83–84 min) and D2 (85–86 min) (Figure 4C). MALDI-TOF MS spectra and analytical HPLC revealed the separation of distinct HPLC peaks in fractions D1 and D2 that were later sequenced as psyle E and caripe 12, respectively (Figure 4C and Supplementary Figure 1).

Amino acid sequences of purified cyclotides derived from fractions C-G (Figure 3B and Supplementary Figure 1) were obtained using an optimized MALDI-based peptidomics approach, termed ‘sequence fragment assembly’ (Hashempour et al., 2013). The trace amounts of cyclotides in fractions A and B precluded their inclusion for sequencing. The overall workflow is illustrated for caripe 8 (Figure 5) and shown for other de novo sequenced cyclotides in the Supplementary Figures 2–7. First, the purified, native cyclotides (Figures 5A,B) were chemically modified to yield S-carbamidomethylation of cysteines. This included the reduction of disulfide bonds with dithiothreitol, and the alkylation of reduced sulfhydryl groups with iodoacetamide, which yielded a mass shift of 348.1 Da (Figure 5C), indicative of the presence of six cysteine residues (as mentioned previously). Afterward, the fully reduced and alkylated peptides were digested with a single enzyme, i.e., trypsin or endo-GluC to produce linear peptide chains amenable to fragmentation by MS/MS. The resulting spectra were analyzed manually by allocating N-terminal b- and C-terminal y-ions (Figures 5D,E). Due to the presence of only one conserved glutamic acid residue in the cyclotide sequence, an endo-GluC digest will usually provide a complete C-N ion series of the linearized precursor ion (caripe 8, 3604.8 Da). On the other hand, tryptic digests of cyclotides derived from HPLC fractionation often results in undistinguishable fragmentation patterns due to multiple enzyme cleavage sites (Arg and Lys), and hence multiple (and sometimes very small fragments, or single amino acids) (Hashempour et al., 2013) (for caripe 8 only the largest precursor with 2770.3 Da provided useful sequence information, Figure 5E). To overcome this problem in de novo cyclotide sequencing, single and double digests combining trypsin, endo-GluC and chymotrypsin were applied to generate smaller fragments with distinct molecular weight (Figure 5F). By combining the annotated sequence information derived from the molecular weight of fragments and alignment with the assigned endo-GluC ion series it was possible to assemble the full cyclotide sequence. In the example of caripe 8, the combination of trypsin/chymotrypsin (1338.5, 1450.6, and 1917.8 Da) and endo-GluC/chymotrypsin provided each three distinct fragments (1001.3, 2129.2, and 2622.5 Da) (Figure 5F). Each full length sequence was generated by alignment of the sequenced endo-GluC and tryptic ion series, with the annotated fragments of the trypsin/chymotrypsin and endo-GluC/chymotrypsin digest. Finally, sequences were confirmed by automated ion fragmentation analysis, chymotrypsin fragmentation pattern, sequence homology, and amino acid analysis (see Materials and Methods section). Applying this approach for sequencing of cyclotides isolated from ipecac fraction A–G led to the identification of seven cyclotides, four of which (caripe 10, 11, 12, and 13) were previously unknown sequences (Table 1). Of the three known cyclotides, two cyclotides, caripe 7 and 8, were only characterized at a transcriptome level by tBLASTn (Wang et al., 2008) mining of the 1kp dataset (Koehbach et al., 2013a), and their molecular structure was here for the first time confirmed at a peptide level (Table 1). The third cyclotide, psyle E has been previously isolated from P. leptothyrsa (Gerlach et al., 2010), possibly indicative of the evolutionary relationship between P. leptothyrsa and C. ipecacuanha.

Sequence analysis of purified C. ipecacuanha cyclotides revealed the existence of the conserved six cysteine residues, the conserved glutamic acid (Glu) residue in loop 1, and the absence of the cis-Pro residue in loop 5, which classifies them to the bracelet sub-family. The seven characterized cyclotides (caripe 7, 8, 10, 11, 12, 13, psyle E) contain the common GES motif in loop 1, the VFIP motif in loop 2 and a serine residue in loop 4 (Table 1). They significantly differ in loops 3 and 5, which are typically recognized to exhibit the highest sequence variability. The only differences in loop 6 are the absence of valine in caripe 7 and caripe 13, and the presence of aspartic acid instead of asparagine in psyle E (Table 1). Although caripe 7 and psyle E share the same molecular weight (3254.5 Da), these cyclotides differ in their loop 3 sequences, i.e., TVTALL (caripe 7) vs. ISSVLG (psyle E). Similarly, caripe 8 and 13 (both have the molecular weight 3237.4 Da), contain the sequences ITAAI and FTSVF in loop 3, respectively. The sequence differences of the co-eluting cyclotides of fraction C (caripe10 and 11) and fraction D (caripe 12 and psyle E) (Figure 4) can be found in loops 3, 5, and 6. Caripe 10 and 11 differ in only two residues – phenylalanine vs. isoleucine (loop 3) and lysine vs. asparagine (loop 5), respectively – resulting in a molecular weight difference of 20 Da (Figure 4 and Table 1). The other set of co-eluting cyclotides from fraction D, caripe 12 and psyle E, differ by 35 Da, which is due to the presence of phenylalanine, isoleucine (loop 3) and asparagine (loop 6) in caripe 12, instead of isoleucine, leucine (loop 3) and aspartic acid (loop 6) in psyle E. Following the detailed sequence analysis of cyclotides isolated from C. ipecacuanha that were able to modulate signaling of the CRF₁R, it was of interest to characterize this effect in pharmacological detail.

We sought to further characterize the inhibitory effect of a purified cyclotide at the CRF₁R. Based on the previous bioactivity fractionation findings, we selected caripe 8 for subsequent pharmacological analysis as this was the most effective inhibitor of CRF₁R-mediated luciferase activity (fraction F; Figure 3C) relative to the other identified cyclotides. To attempt to disambiguate the effects of cytotoxicity and non-receptor-linked convergent signaling pathways from the previous luciferase bioassay, we sought to examine CRF₁R-mediated cAMP accumulation. The proximal nature of cAMP generation relative to receptor activation enables more direct assessment of CRF₁R pharmacological modulation. Similar to the luciferase assay, caripe 8 did not exhibit any intrinsic activity in the cAMP assay. Additionally, no effect by the cyclotide on cAMP accumulation was observed when in combination with submaximal agonist concentrations of forskolin. The initial titration of 2.6, 26, and 260 nM of the cyclotide with an EC₅₀ of CRF revealed modest concentration-dependent inhibition of CRF₁R-mediated cAMP responses (Figure 6A), and was observed to reach a threshold (data not shown). Further exploration of this antagonism determined a small (approximately 4.5-fold) reduction of CRF potency by 260 nM caripe 8, but found a lack of effect on the agonist’s maximal response (Figure 6B). In contrast, caripe 8 did not inhibit arginine-vasopressin-, or forskolin-stimulated cAMP responses at the vasopressin V₂-receptor (Supplementary Figure 8), suggesting a CRF₁R-specific mode-of-action.

Carapichea ipecacuanha root powder (cat. no. 66804) was purchased from Alfred Galke GmbH (Bad Grund, Germany). The plant material (200 g) was extracted with 1 L of methanol/dichloromethane, 1:1 (v/v) for 18–24 h by maceration under continuous agitation at 25°C. After filtration 0.5 volume of ddH₂O was added to the extract, and the methanol/water phase, which contained the cyclotides, was obtained by liquid/liquid phase separation. This aqueous mixture was further pre-purified by C₁₈ solid-phase extraction. The dried, crude extract was dissolved in 10% methanol/90% ddH₂O (v/v) and then loaded onto C₁₈ material ZEOprep 60 Å, irregular 40–64 μm (Zeochem, Uetikon, Switzerland) that had been activated with methanol and equilibrated with solvent A (100% ddH₂O/0.1% trifluoroacetic acid, v/v). After washing with 10% solvent B (90% acetonitrile/10% ddH₂O/0.08% trifluoroacetic acid, v/v/v) it was eluted with 80% solvent B to separate the cyclotide-containing fraction from polar compounds. This pre-purified extract, containing cyclotides, is referred to as ‘ipecac extract.’ Following initial analysis of the pre-purified ipecac extract, cyclotides were enriched by a second cycle of solid phase extraction including washing of the extract with 30% solvent B and elution with 80% solvent B. The resulting fraction is referred to as ‘cyclotide-enriched ipecac fraction.’

The dried ipecac extract or cyclotide-enriched ipecac fraction were dissolved in solvent A for liquid chromatography. The mobile phase for all HPLC analysis and purifications consisted of solvent A (100% ddH₂O/0.1% trifluoroacetic acid, v/v) and solvent B (90% acetonitrile/10% ddH₂O/0.08% trifluoroacetic acid, v/v/v) and separations were performed on a Dionex 3000 LC unit (Dionex, Amsterdam, The Netherlands), as published earlier (Hellinger et al., 2015a). Analytical RP-HPLC was carried out using a Kromasil C₁₈ column (250 mm × 4.6 mm, 5 μm, 100 Å; dichrom GmbH, Marl, Germany) at a flow rate of 1 mL min^-1, and peptides were separated with linear gradients of solvent B between 15 and 65% at 0.66% min^-1, including pre- and post-gradient equilibration steps. Cyclotide fractionation and purification was performed at preparative and semi-preparative scale using a Phenomenex Jupiter C₁₈ column (250 mm × 21.2 mm, 10 μm, 300 Å; Phenomenex, Aschaffenburg, Germany) and a Kromasil C₁₈ column (250 mm × 10 mm, 5 μm, 100 Å). The flow rates were set to 8 and 3 mL min^-1, respectively, and the cyclotides were separated with linear gradients of solvent B between 15 and 65% at 0.4% min^-1 (preparative) and 0.57% min^-1 (semi-preparative), including pre- and post-gradient equilibration steps. Elution profile was monitored via UV absorbance at 214, 254, and 280 nm. Cyclotide fractions were collected manually according to their absorbance 214 nm.

Analysis of peptide samples were performed by MS using a MALDI-TOF/TOF 4800 Analyser (AB Sciex, Framingham, MA, United States) operated in reflector positive ion mode acquiring between 2000 and 10000 total shots per spectrum with a laser intensity set between 3200 and 3800. Samples for MS experiments were prepared using α-cyano-hydroxyl-cinnamic acid as matrix (Sigma–Aldrich, St. Louis, MO, United States) in ddH₂O/acetonitrile/trifluoroacetic acid, 50/50/0.1% (v/v/v) at a concentration of 5 mg mL^-1. An aliquot of each sample (0.5 μL) was mixed with of matrix solution (3 μL) and spotted directly onto the MALDI target plate. Spectra were processed and analyzed using the Data Explorer Software^TM (AB Sciex).

Reduction of disulfide bonds was performed by dithiothreitol (Sigma–Aldrich) using 2 μL of a freshly prepared 0.2 M solution (prepared in 0.1 M NH₄HCO₃, pH 8.2) that was added to 20 μL of extract, cyclotide-enriched fraction or purified cyclotide dissolved at a concentration of 5 mg mL^-1 in 0.1 M NH₄HCO₃ buffer (pH 8.2). The reduction mixture was allowed to incubate for 30 min at 37°C. Afterward, the reduced samples were carbamidomethylated for 10 min at 23°C in the dark by adding 4 μL of 100 mM iodoacetamide (prepared in 0.1 M NH₄HCO₃, pH 8.2 by briefly heating to 65°C for 1 min) to the mixture. The alkylation reaction was stopped by adding 1 μL of 10 mM dithiothreitol. Samples were used for MS analysis or enzymatic digest in preparation for peptide sequencing. For digestion the purified cyclotides (after reduction and alkylation) was performed by adding 2 μL of trypsin, endo-GluC, or chymotrypsin (0.1–0.5 μg μL^-1, Sigma–Aldrich). All digests were incubated at 37°C for 3–16 h, quenched with concentrated acetic acid (final concentration 3%) and stored at -20°C until further analysis.

Tandem mass spectra were acquired using a laser energy of 1 kV, with or without the use of collision-induced dissociation. The spectra were examined and sequenced based on assignment of the N-terminal b-ion and C-terminal y-ion series in combination with automated sequence analysis using the Data Explorer Software. The disulfide connectivity of the conserved cysteines (C_I-IV, C_II-V, and C_{III-V I}) and the isobaric amino acid Leu and Ile were assigned based on homology with published sequences, and/or chymotrypsin fragmentation pattern. In addition the amino acid compositions of novel, purified peptides (caripe 7, 8, and 10; 100–200 μg) were confirmed with high sensitivity amino acid analysis using gas hydrolysis (Australian Proteome Analysis Facility, Sydney, NSW, Australia).

HEK293 cells were co-transfected with CRF₁R plasmid and firefly luciferase containing plasmid pGL4.29 luc2P (Promega GmbH, Mannheim, Germany) expressing the downstream DNA binding elements of GPCR activation (CRE for Gα_s-coupled signaling). Using CaPO₄ transfection (Koehbach et al., 2013b), 2.4 × 10⁶ of HEK293 cells per 10 cm dish were prepared and cultured in Dulbecco’s modified Eagle’s medium (PAA, Pasching, Austria) supplemented with 10% fetal bovine serum at 37°C with a humidified atmosphere at 5% CO₂ for 16 h. Then 6 μL of each needed plasmid (CRF₁R and CRE for Gα_s-coupled signaling) with a concentration of 1 μg μL^-1, was mixed with 287.4 μL H₂O + CaCl₂ (250 μL H₂O + 37.4 μL CaCl₂) and added to 300 μL of HEPES-buffered saline buffer. After 20 min, the solution was added to the medium and the cells were incubated for 6 h at 37°C. Cells were then seeded into 96-well plates (5 × 10⁴ cells per well) and incubated with different concentrations of cyclotide-enriched fractions with and without 50 nM endogenous agonist CRF in white medium supplemented with 10% fetal calf serum (Gibco, Gaithersburg, MD, United States) plus antibiotics. Following 6 h of incubation at 37°C, medium was removed and then cells were lysed using the luciferase cell culture lysis reagent (Promega GmbH). The luminescence and fluorescence intensities were measured via the Synergy H4 microplate reader (Biotek Instruments, Winooski, VM, United States). The measured luciferase counts were corrected using the fluorescent intensity for the number of cells per well and normalized to the maximum activation of endogenous CRF agonist of the CRF₁R.

Molar extinction coefficients (𝜀 at 280 nm) of novel cyclotides were determined according to the formula 𝜀₂₈₀ = nC^∗120 + nW^∗5690 + nY^∗1280 [M^-1 cm^-1] (where n is the number of residues). The concentration of reconstituted purified peptides was calculated by the Beer-Lambert law based on absorbance measurements at 280 nm using the Nanodrop 1000 (Thermo Fisher Scientific, Waltham, MA, United States). Accordingly the molar extinction coefficient (𝜀₂₈₀) of caripe 8 was calculated to 2000 M^-1 cm^-1.

Receptor-mediated cAMP accumulation was measured in HEK293 cells transiently transfected with the human CRF₁R (with C-terminal GFP tag) using the LANCE^® Ultra cAMP Detection kit (PerkinElmer, Waltham, MA, United States). To prepare for experiments, HEK293 cells were grown to confluence in a 6-well plate, followed by transfection with 2.5 μg of plasmid DNA per well using the Lipofectamine^® 2000 transfection reagent (Thermo Fisher Scientific). After 4–6 h post-transfection, the cell media was replaced, followed by dissociation of cells via trituration and division of the cell suspension to avoid overcrowding over the subsequent 48 h of incubation at 37°C and 5% CO₂. On the day of assay, cells were dissociated with phosphate buffered saline- (PBS-)EDTA solution (137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 1.8 mM KH₂PO₄, 1 mM EDTA) and re-suspended in stimulation buffer (1.26 mM CaCl₂, 0.49 mM MgCl₂, 0.41 mM MgSO₄, 5.33 mM KCl, 0.44 mM KH₂PO₄, 4.17 mM NaHCO₃, 138 mM NaCl, 0.34 mM Na₂HPO₄, 5.56 mM D-glucose, 5 mM HEPES, 0.5 mM IBMX, 0.1% BSA; pH 7.4). Stimulation of cAMP accumulation via the human CRF₁R and the human V₂R used 500 cells and 300 cells (expressing these receptors) per well, respectively, in 384-well white Cellstar^® plates (Greiner Bio-One, Kremsmuenster, Austria). Cells were incubated with the respective cognate receptor agonist peptides, in the absence and presence of cyclotide, for 30 min at 23°C in a final volume of 10 μL of stimulation buffer. Termination of cAMP accumulation was performed with the addition of 5 μL of Eu-cAMP tracer solution and 5 μL of ULight-anti-cAMP antibody solution, each prepared in detection buffer, followed by incubation of the plate for 1 h at 23°C with the lid on. After incubation, the plate was read in a Flexstation 3 plate reader (Molecular Devices, Sunnyvale, CA, United States).

Data from cellular assays were analyzed using GraphPad Prism 7.01 (GraphPad Software, San Diego, CA, United States). Concentration-response curve data were fitted to the three-parameter logistic equation to derive estimates of potency (LogIC₅₀, luciferase; LogEC₅₀, cAMP) and efficacy (E_max), relative to the maximal agonist control response. CRF LogEC₅₀ values in the absence and presence of caripe 8 were statistically compared via extra sum-of-squares F test analysis. Other statistical analyses were performed as appropriate using one-way analysis of variance (ANOVA) with Bonferroni’s post-test, and significance taken as p < 0.05.