Discovery and development of macrocyclic peptide modulators of the cannabinoid 2 receptor

Abstract

The cannabinoid type 2 receptor (CB₂R), a G protein–coupled receptor, is an important regulator of immune cell function and a promising target to treat chronic inflammation and fibrosis. While CB₂R is typically targeted by small molecules, including endo-, phyto-, and synthetic cannabinoids, peptides–owing to their size–may offer a different interaction space to facilitate differential interactions with the receptor. Here, we explore plant-derived cyclic cystine-knot peptides as ligands of the CB₂R. Cyclotides are known for their exceptional biochemical stability. Recently, they gained attention as G protein–coupled receptor modulators and as templates for designing peptide ligands with improved pharmacokinetic properties over linear peptides. Cyclotide-based ligands for CB₂R were profiled based on a peptide-enriched extract library comprising nine plants. Employing pharmacology-guided fractionation and peptidomics, we identified the cyclotide vodo-C1 from sweet violet (Viola odorata) as a full agonist of CB₂R with an affinity (K_i) of 1 μM and a potency (EC₅₀) of 8 μM. Leveraging deep learning networks, we verified the structural topology of vodo-C1 and modeled its molecular volume in comparison to the CB₂R ligand binding pocket. In a fragment-based approach, we designed and characterized vodo-C1-based bicyclic peptides (vBCL1-4), aiming to reduce size and improve potency. Opposite to vodo-C1, the vBCL peptides lacked the ability to activate the receptor but acted as negative allosteric modulators or neutral antagonists of CB₂R. This study introduces a macrocyclic peptide phytocannabinoid, which served as a template for the development of synthetic CB₂R peptide modulators. These findings offer opportunities for future peptide-based probe and drug development at cannabinoid receptors.

The endocannabinoid system comprises the endogenous arachidonic acid-derived ligands anandamide and 2-arachidonoylglycerol (2-AG), enzymes facilitating endocannabinoid transport and metabolism, and importantly, the cannabinoid type 1 and 2 receptors (CB₁R and CB₂R). These G protein-coupled receptors (GPCRs) play a fundamental role in maintaining homeostasis in vertebrates across various physiological processes. Due to its expression in peripheral lymphoid tissues and cells of the immune system, the CB₂R emerged as a potential therapeutic target for numerous diseases including autoimmune disorders, metabolic conditions, cancer, and chronic inflammatory disorders (1, 2). CB₂R couples to inhibitory G_i/o proteins, which trigger the activation of downstream effector pathways, such as inhibition of cAMP accumulation, activation of extracellular signal-regulated kinase 1/2, and the recruitment of β-arrestins to the receptor (3). However, it is currently unknown which signaling cascades are most relevant for therapeutic benefit. Although considerable efforts have been made to develop CB₂R-selective therapeutics, none have yet reached the market, being unable to meet efficacy end points within clinical evaluation. High lipophilicity and poor target engagement of small molecule cannabinoid ligands are usually associated with high toxicity and unwanted side effects, whereas applying lower, tolerated doses results in lack of clinical efficacy (4). Addressing the current challenges of cannabinoid modulators and chemical probes is crucial for advancing our understanding of CB₂R pharmacology and unlocking the therapeutic potential of this receptor (5). Recently, peptides have gained attention as modulators of cannabinoid receptors, exemplified by the discovery of endogenous hemoglobin-derived peptide allosteric modulators of cannabinoid receptors (6, 7, 8). As alternative to small molecule drugs, peptides offer potentially enhanced safety profiles (9, 10), combining the potency and selectivity of larger proteins and antibodies with more efficient and cost-effective production (10). The increasing discoveries of peptide GPCR ligands in recent years have primarily been driven by advancements in computational biology (11), structure-based design (12), virtual library screening (13), and rational design inspired by natural scaffolds (14, 15, 16).

While Cannabis sativa L. produces the tetrahydrocannabinol-type phytocannabinoids, also other cannabinoid receptor ligands from plants have been identified (17). Notably, plants offer an extensive chemical space of bioactive peptides, with cyclotides evolving as innovative candidates for GPCR ligand discovery and design (9, 15, 18, 19, 20). These plant-derived peptides contain the structural feature known as cyclic cystine-knot motif (21), which sparked their use templates for the design of stable peptide-based therapeutics with an improved pharmacokinetic profile (18, 22, 23). Building on convenient and robust technologies that allow their (i) analytical characterization using peptidomics (24, 25), (ii) automated access to correctly folded peptides via solid-phase chemistry (26), and (iii) access to rapid plate-based pharmacological screening assays (19), in this study, we aimed to explore cyclotide-containing plants to identify modulators of the CB₂R.

Using a pharmacology-guided fractionation workflow comprising nine peptide-enriched plant extracts, we identified a cyclotide from sweet violet (Viola odorata) that displaced an orthosteric small molecule radioligand ([³H]-CP55940) from the binding site of CB₂R, and activated the receptor, as determined by second messenger quantification. Based on the unique three-dimensional fold of this peptide, we designed and synthesized bicyclic analogs and provided comprehensive pharmacological characterization of these molecules at the CB₂R. Given their stability, these peptide ligands arising from this study will serve as novel research tools for dissecting cannabinoid receptor pharmacology and inspire future drug development.

Results

Screening of a cysteine-rich peptide library to discover CB₂R ligands

Driven by the concept of nature-derived peptides as source for identification of new chemical scaffolds to modulate GPCR signaling (9, 18), we prepared a peptide-enriched plant extract library comprising nine plant species previously recognized to express a suite of cysteine-rich peptides. This included cyclotides from Carapichea ipecacuanha, Palicourea tomentosa, V. odorata, and Viola tricolor (27, 28), knottin-like peptides from Bryonia alba (29), Salix alba, and Strophanthus kombe, as well as short protease inhibitor-like peptides from Helianthus annuus (30) and Citrus limon. Using chemical solvent and solid phase extraction, we generated a diverse library containing myriad of unknown or previously identified peptides and documented their mass signals in the range of 2500 to 4000 Da with MALDI-TOF mass spectrometry (MS) analysis (Fig. S1). First, we performed displacement of [³H]-CP55940 binding at human CB₂R stably expressed in CHO-K1 membranes by WIN55,212-2 to validate assay conditions (Fig. S2). Subsequently, we screened the peptide library in radioligand displacement binding assays at the CB₂R (Fig. 1A). Cyclotide-containing plant extracts were able to displace the CB₂R orthosteric small molecule radioligand agonist [³H]-CP55940, and the sweet violet extract (V. odorata) exhibited the most pronounced binding effect (Fig. 1, A and B). To identify and isolate a CB₂R cyclotide ligand from V. odorata, we utilized a pharmacology-guided screening and purification approach. Preparative reversed-phase (RP)-HPLC of V. odorata extract yielded six cyclotide-enriched fractions A-F (Figs. 1C and S3, A–F), which were assayed in radioligand binding experiments at CB₂R. Cyclotide-rich fractions A-D displayed none to moderate displacement of [³H]-CP55940 (70–130% of radioligand bound), whereas fractions E and F (Fig. S3, E and F), exhibited the strongest ability to displace radioligand from the CB₂R (12% and 2% of radioligand bound, respectively) (Fig. 1D).

Figure 1.

Screening of peptide-enriched plant extracts on CB₂R.A, binding assays utilizing radioligand displacement of [³H]-CP55940 (0.4 nM, dark gray bar) by peptide-enriched plant extracts, that is, Bryonia alba, Carapichea ipecacuanha, Citrus limon, Helianthus annuus, Palicourea tomentosa, Salix alba, Strophanthus kombe, Viola odorata, and Viola tricolor (300 μg/ml, light gray bars; extract of V. odorata was used for further purification, red bar) using human CB₂R-containing membrane preparations. WIN55,212-2 (10 nM, white bar) was used as positive control. B, mass spectrometry analysis (MALDI-TOF) of V. odorata extract with the major mass signals (>20% base peak intensity) shown as monoisotopic masses [M + H]⁺. Known peptide masses (±1 m/z) were labeled with the corresponding cyclotide name published in CyBase (31). C, preparative RP-HPLC chromatogram of the peptide extract from V. odorata with the peptide-rich fractions (denoted as A–F) separated by dotted lines. D, radioligand displacement binding of [³H]-CP55940 (0.4 nM, dark gray bar) by semipurified peptide fractions A–F (300 μg/ml, light gray bars; fraction F was used for further purification, red bar) using human CB₂R containing membrane preparations. Specific binding was obtained by subtracting of nonspecific from total binding. Data are presented with individual data points and bars as mean ± SD (n = 3) and are normalized to the fraction of maximum bound radioligand, which refers to an average of 4500 to 5000 fmoles/mg protein for CB₂R. CB₂R, cannabinoid type 2 receptor; RP, reversed-phase.

Identification of the CB₂R ligand vodo-C1 from V. odorata

We next examined the cyclotide content of fraction F, which exhibited the strongest displacement of radioligand at CB₂R. MALDI-TOF and analytical HPLC analysis of fraction F revealed the presence of nine known cyclotides (24, 31) (Fig. S3F, Table S1) and one previously unidentified peptide with a molecular weight of 3432.2 Da. Since V. odorata is a well-documented cyclotide-expressing plant (32) we assumed the unknown peptide, which was named vodo-C1, is also a cyclotide. Hence, we first applied a chemical derivatization approach to determine the cysteine content of vodo-C1.

Chemical reduction of the (putative) disulfide bonds by DTT and S-carbamidomethylation of cysteines with iodoacetamide revealed a shift from 3432.1 Da (native) to 3438.1 Da (reduced) and 3780.2 Da (alkylated), resulting in mass difference of +6 Da and +348 Da, respectively, which corresponds to a peptide containing six cysteine residues (i.e. three disulfide bonds) (Fig. 2A). To elucidate the amino acid sequence of this peptide and to determine if the backbone is cyclized, the S-carbamidomethylated peptide was enzymatically processed with endoprotease GluC (EndoGluC), trypsin and chymotrypsin and subsequently analyzed by mass spectrometry. EndoGluC proteolytic digest revealed a mass peak of 3798.4 Da, which corresponds to an increase of +18.2 Da due to hydrolysis of the S-carbamidomethylated peptide. This suggested “ring-opening” of the peptide by hydrolysis, which is a main feature of cyclotides (27).

Figure 2.

Identification and amino acid sequence elucidation of cyclic disulfide-rich peptide from Viola odorata.A, MALDI-TOF mass spectrum of native (3432.1 Da), reduced (3438.1 Da) and S-carbamidomethylated peptide (3780.2 Da) revealed mass shifts of +6.0 Da after and +348.1 Da, observed after DTT-reduction and S-carbamidomethylation, respectively. B, MALDI-TOF spectrum of S-carbamidomethylated full-length vodo-C1 peptide after digestion with EndoGluC. All labeled peaks refer to monoisotopic masses [M + H]⁺. In total, two fragments were detected indicating cleavage of the peptide at two distinct positions. C, MS/MS spectrum of the mass 1252.5 Da and (D) the mass 2565.6 Da precursor with identified fragment b- and y-ions labeled (monoisotopic [M + H]⁺). MS, mass spectrometry.

Two additional fragments have been identified in this digest with masses of 1252.4 Da and 2564.9 Da, indicating additional cleavage of the linear precursor at Glu residues (Fig. 2B). Each of the tryptic and chymotryptic digests yielded two cleavage products of the linear precursor, with masses of 1652.6 Da and 2164.8 Da, and 628.2 Da and 3189.2 Da, respectively (Figs. S4A and S5A). The above identified proteolytic cleavage products were subjected to collision-induced fragmentation by MALDI-TOF MS/MS. The resulting MS/MS spectra were manually annotated for their N-terminal b- and C-terminal y-ion series (Figs. 2, C and D, S4, B and C and S5, B and C). The integration of sequence data from these different proteolytic experiments enabled a complete assignment of the peptide sequence, which is cyclo-GDPLPCGETCFTGKCYSETIGCTCEWPICTKN. In addition, fragment analysis of the trypsin digest allowed distinction between isobaric residues Gln/Lys (Fig. S4, B and C), and the chymotrypsin digest was used to distinguish between isobaric Leu/Ile (Fig. S5, B and C). Finally, the sequence of vodo-C1 was confirmed by high-sensitivity amino acid analysis and sequence similarity analysis (31) (Fig. S6, Table S2).

Synthesis of vodo-C1 and pharmacological characterization reveals CB₂R full agonism

Following de novo sequencing, vodo-C1 was chemically synthesized for pharmacological characterization. A combination approach using fluorenylmethoxycarbonyl (Fmoc) chemistry, followed by N-to-C-terminal backbone cyclization using peptide hydrazide as thioester precursor and subsequent oxidative folding yielded the cyclotide vodo-C1 (Fig. S7, A–C). Its native disulfide configuration was confirmed by analytical HPLC coelution analysis in that synthetic and plant-extracted vodo-C1 produced a single peak at 38.9 min (Fig. 3A). Subsequently, vodo-C1 was subjected to concentration-dependent binding studies and second messenger quantification at the CB₂R. Compared to CP55940, a full reference agonist of the CB₂R (3), vodo-C1 displaced [³H]-CP55940 in a concentration-dependent manner with a K_i value of 0.9 ± 0.2 μM (Fig. 3B). Functional cAMP assay revealed that vodo-C1 activated CB₂R, with an E_max of 102.8 ± 7.1% and an EC₅₀ of 7.8 ± 1.7 μM (Fig. 3C). These data suggest that vodo-C1 is a full agonist of CB₂R and more importantly, it is the first identified peptide agonist of CB₂R.

Figure 3.

Pharmacology of synthetic vodo-C1 cyclotide.A, quality control of synthesized vodo-C1 (black signal) analyzed by RP-HPLC (purity > 95%) with retention time of 38.8 min and coelution experiment of native and synthetic vodo-C1 (ratio 1:1; red dashed signal) showing one single peak at 38.9 min. B, radioligand displacement of [³H]-CP55940 (0.3 nM) by synthetic vodo-C1 (red circles) using human CB₂R containing membrane preparations led to a calculated affinity (K_i) of 0.9 ± 0.2 μM. CP55940 was used as a positive control (black circles). Data are presented as mean ± SD and are normalized to the percentage of maximum binding, which refers to an average of 4500 to 5000 fmoles/mg protein for CB₂R (n = 3). C, concentration-dependent cAMP inhibition following full receptor activation by synthetic vodo-C1 (red circles) in CHO-K1 cells stably expressing the human CB₂R with EC₅₀ value of 7.8 ± 1.7 μM and E_max of 102.8 ± 7.1%. CP55940 was used as a positive control (black circles; n = 3). CB₂R, cannabinoid type 2 receptor; RP, reversed-phase.

Vodo-C1-inspired design of cannabinoid receptor ligands

To better understand the interaction of the cyclotide ligand with the receptor, we modeled vodo-C1 using AlphaFold (33), combined with workflow for structure prediction and design of cyclic peptides (34). The cyclotide has a typical Möbius fold given the presence of a cis-prolyl peptide bond in loop 5, which causes a twist in the circular backbone (28) and a cyclic peptide backbone with three disulfide bonds (connectivity: C_I-C_IV, C_II-C_V, and C_III-C_VI) (Fig. 4A). Besides one conserved glutamic acid residue in loop 1 (35), there are an additional two Glu residues in loops 3 and 5. This is an unusual feature since the majority of cyclotides contain one conserved Glu residue (27). Hitherto, only five cyclotides have been identified with three Glu residues, including cliotide T2 (36), tricyclon B (37), viba 14 (38), hyfl B, and hyfl C (39), making vodo-C1 one of the cyclotides with this unique amino acid composition. Overall, the model displayed predicted local distance difference test-score with high confidence (predicted local distance difference test = 0.947) and a predicted aligned error between 0 and 15 for all residues (Fig. S8, A and B). Vodo-C1 aligned with the prototype cyclotide structure of kalata B1 with a RMSD-value of 0.648 Å (Fig. 4B). Residues in loop 2 (Phe11) and loop 5 (Thr26 and Ile28) form a hydrophobic patch with a calculated hydrophobicity index of 2.77 (according to the hydrophobicity scale by Eisenberg (40)) (Fig. 4C). Furthermore, we calculated the molecular volumes of vodo-C1 in comparison to kalata B1, and other cannabinoid CB₂R ligands (Table S3). Vodo-C1 has a volume of 3516 Å³, whereas the CB₂R binding pocket has a calculated volume ranging from approximately 415 to 447 Å³ (41).

Figure 4.

Structure prediction of native vodo-C1 using AfCycDesign.A, a cartoon representation of predicted structure of cyclotide vodo-C1. The sulfur atoms of the six cysteine residues are highlighted in yellow. B, a structural alignment of kalata-B1 (PDB ID: 1NB1) and vodo-C1 using PyMol. An RMSD-value of 0.648 Å was calculated for the alignment of the two peptides. C, a surface representation of vodo-C1. A hydrophobicity scale according to Eisenberg (32) was applied to the model. Hydrophobic residues are highlighted in red, whereas hydrophilic residues are shown in white. Hydrophobic residues (L4, F11, I20, W26, and I28) are highlighted in white. D, sequence alignment of vodo-C1 with kalata B1. Conserved cysteine residues (I-IV) are highlighted in yellow and labeled with Roman numbers (in red) above the alignment, whereas cyclotide loops (1–6) connecting different cysteine residues are labeled with Arabic numbers (in blue) underneath the alignment. PDB, Protein Data Bank.

Given the bulky size of cyclotides, we hypothesized a smaller surface area and volume of vodo-C1-inspired shorter peptides would enable better penetration of the binding pocket of the receptor and exhibit increased affinity and/or potency (14). Therefore, in a fragment-based approach we designed four smaller, bicyclic peptides using the native sequence of vodo-C1 as template termed vodo-C1-inspired bicyclic loop peptides 1 to 4 (vBCL1-4) (Fig. 5A). We synthesized linear peptides, each composed of two native vodo-C1 loops and three cysteine residues: vBCL1 (loops 1 and 2, containing 10 residues), vBCL2 (loops 2 and 3, containing 13 residues), vBCL3 (loops 5 and 6, containing 15 residues), and vBCL4 (loops 6 and 1, containing 14 residues). Loop 4 was excluded since it comprises a single amino acid residue. Cyclization was achieved by using the 1,3,5-tris-(bromomethyl)-benzene (42), which contains three thio-reactive groups able to couple with three cysteine moieties of each of linear peptides. (Figs. 5B and S9, A–D).

Figure 5.

Schematic illustration of design and synthesis strategy of novel bicyclic peptides to target CB₂R.A, fragment-based design of bicyclic peptides vBCL1-4, each comprising two loops of native vodo-C1 cyclotide. B, peptide preparation via solid phase peptide synthesis and cyclization by coupling the Heinis reagent 1,3,5-tris-(bromomethyl)-benzene, containing three thio-reactive moieties that reacted with the three thiol groups of the cysteine residues of each linear peptide. CB₂R, cannabinoid type 2 receptor; vBCL, vodo-C1-inspired bicyclic loop.

Next, we determined the molecular pharmacological properties of vBCL peptides to obtain information regarding their affinity and functional effects at CB₂R. The peptides were measured in two-point displacement radioligand assays at CB₂R. vBCL1 and 2 (10 μM) exhibited weak displacement of [³H]-CP55940 from CB₂R (75–90% radioligand bound), whereas vBCL3 and 4 (100 nM and 10 μM, respectively) did not bind to the receptor (Fig. 6A). The functional cAMP assays confirmed that all four vBCLs did not activate the human CB₂R at concentrations up to 30 μM (Fig. 6B). To determine whether the designed vBCL peptides and the parent molecule vodo-C1 are able to bind to CB₁R, we performed single-point displacement radioligand assays on HEK293 membranes transiently expressing rat CB₁R. Fraction F (300 μg/ml) exhibited weak displacement of [³H]-CP55940 (∼80% of radioligand bound) at the CB₁R, whereas vodo-C1 (30 μM) showed no displacement of radioligand (∼130% of radioligand bound), indicating selectivity for CB₂R (Fig. S10A). Furthermore, the vBCLs did not displace [³H]-CP55940 from CB₁R at concentrations of 10 μM (Fig. S10B).

Figure 6.

Pharmacology of vBCL1-4 at the CB₂R.A, two-point radioligand displacement assay of peptides at the CB₂R. Radioligand [³H]-CP55940 (0.5 nM; dark gray bar) mixed with cold CP55940 (4.5 nM) with or without 10 μM (gray bars) or 100 nM (dotted white bars) of peptides. WIN55,212-2 (1 μM; white bar) was used as a positive control. Data are presented as individual data points and bars as mean ± SD (n = 3) and are normalized to the percentage of maximum binding, which refers to an average of 4500 to 5000 fmoles/mg protein for CB₂R. B, inhibition of cAMP accumulation by bicyclic peptides at the CB₂R was measured in stable CHO-K1 cells using CP55940 as reference ligand (n = 3). CB₂R, cannabinoid type 2 receptor; vBCL, vodo-C1-inspired bicyclic loop.

Cyclotide-derived bicyclic peptides vBCL1-4 are distinct CB₂R modulators with allosteric properties

Given the observed lack of strong radioligand displacement and activation of the CB₂R, we screened vBCLs for allosteric properties at CB₂R using functional second messenger assays. By measuring CB₂R-mediated inhibition of cellular cAMP production, we demonstrated that 10 μM vBCL2 and vBCL4 decreased the potency (EC₅₀) of the full agonist CP55940 by approximately 5-fold from 15 nM to 80 nM and 84 nM, respectively. In contrast, the shift in CP55940 potency with 10 μM of vBCL1 or vBCL3 was weaker (∼2-fold for each). However, it is important to note that vBCL3 and vBCL4 both led to reduction in basal response. Additionally, vBCL3 also decreased the efficacy of CP55940 by ∼40% (E_max = 61%) (Fig. 7A, Table S4). Based on their effects on CP55940 activity, the distinct effects of peptides vBCL2-4 were analyzed in more detail.

Figure 7.

Allosteric effects of vBCL1-4 on CP55940 concentration-response curves at the CB₂R.A, functional cAMP assay of 10 μM vBCL1 (red circles), vBCL2 (green circles), vBCL3 (blue circles), and vBCL4 (violet circles) coincubated with increasing concentrations of CP55940 agonist in CHO-K1 cells stably expressing CB₂R. Calculated fold change of CP55940 EC₅₀ for each vBCL is presented as table inset. Dose-dependent allosteric modulation of CP55940 concentration-response curve in presence of 1 μM (inverted triangles), 3 μM (diamonds), and 10 μM (squares) of (B) vBCL2 (green), (C) vBCL3 (blue), and (D) vBCL4 (violet). Data are normalized to percentage of maximal activation, detected at the highest CP55940 concentration, and are shown as mean ± SD of 3 to 4 independent experiments. CB₂R, cannabinoid type 2 receptor; vBCL, vodo-C1-inspired bicyclic loop.

To examine whether these effects observed are concentration-dependent, we tested concentrations of 1 μM, 3 μM, and 10 μM of vBCL2-4 (Fig. 7, B–D, Table S5). The observed dextral shift of the CP55940 concentration-response curve by increasing concentrations of vBCL2 was best fit using the Gaddum/Schild EC₅₀ nonlinear regression model indicative of competitive antagonism with an estimated functional affinity of ∼1.6 μM (pA2 = 5.8 ± 0.3; Table S6). This agrees with the partial displacement of [³H]-CP55940 induced by vBCL2 (Fig. 6A). vBCL4 behaved similar to vBCL2, inducing a decrease of the CP55940 potency (Fig. 7D, Table S5). However, this concentration-dependent inhibition appeared saturable and was best fit using the operational model of allosterism, yielding an estimated affinity of ∼309 nM (pK_B = 6.5 ± 1.2; Table 1).

Table 1.

Analysis of vBCL4 concentration-dependent inhibition of CP55940-, WIN55,212-, or 2-AG - mediated cAMP formation at CB₂R

Agonist	vBCL4
	log(αβ)a	pKBab	pA2c	Slopec
CP55940	−0.63 ± 0.4	6.5 ± 1.2 (309 nM)	5.7 ± 2.0 (2.0 μM)	0.3 ± 0.3
WIN55,212-2	−1.03 ± 0.8	6.2 ± 0.4 (646 nM)	n.d.	n.d.
2-AG	−1.26 ± 0.4	6.5 ± 0.3 (300 nM)	n.d.	n.d.

Data are presented as means ± SD of 3 to 4 individual experiments.

^aValues estimated via operational model of allosterism and agonism (Equation 1) from data presented in Figures 7D and 8, E and F.

^bnegative logarithm of molar affinity estimate; statistical significance was analyzed using an unpaired t test (p > 0.05).

^cValues estimated via Gaddum/Schild equation.

Co-incubation of vBCL3 (1 μM and 3 μM) marginally altered the potency of CP55950 but induced downward shifts of the CP55940 concentration-response curve reducing the efficacy from 100% to ∼80% and 86%, respectively (Fig. 7C, Table S5). Importantly, the basal activity was also reduced, and therefore operational model analysis was not possible. To validate that the observed reduction of basal activity mediated by vBCL3 is not due to potential intrinsic inverse agonism, we studied the effect of vBCL3 on CP55940 activity in the absence of forskolin (a known activator of adenylyl cyclase (43)). CB₂R is known for its constitutive activity, which leads to the basal inhibition of adenylyl cyclase (44, 45). We hypothesized that in the absence of forskolin, inverse agonism would reduce the constitutive activity of CB₂R and enhance the production of cAMP. Firstly, we confirmed that the assay system is sensitive to detect inverse agonism by measuring effects of varying concentrations of inverse agonist SR144528 (45), which reduced the basal levels of CB₂R activity (Fig. S11A). In the absence of forskolin, neither CP55940 alone, nor CP55940 co-incubated with vBCL3 yielded any effect on cellular cAMP levels (Fig. S11B), concluding that vBCL3 lacks intrinsic inverse agonist activity.

Lastly, possible probe-dependent effects of all four vBCL peptide analogs were characterized, by extending the functional assays with the potent synthetic CB₂R full agonist WIN55,212-2 and the endogenous 2-arachydonylglycerol (2-AG) (3). Specifically, co-incubation of vBCL1 and vBCL3 peptide at 10 μM with WIN55,212-2 or 2-AG, respectively, resulted in EC₅₀ shifts of ∼6-fold (Fig. 8, A and B, Table S4). Interestingly, vBCL3 induced both rightward and downward shifts of the WIN55,212-2 concentration-response curve, modulating the agonist potency (EC₅₀) from 1.5 nM to 9.4 nM and the maximal response (E_max) from 100% to ∼54%. Furthermore, vBCL3 also reduced the basal activity of WIN55,212-2. On the other hand, vBCL2 and vBCL4 had larger effects on agonist potencies. Co-incubation of vBCL2 and vBCL4 peptide at 10 μM with 2-AG yielded EC₅₀ shifts by ∼9 to 12 fold, from 480 nM to 4.4 μM and 5.9 μM, respectively. Also, treatment with 10 μM vBCL2 and vBCL4 shifted the EC₅₀ of WIN55,212-2 by approximately 29-fold and 75-fold, from 1.5 nM to 47 nM and 120 nM, respectively (Fig. 8, A and B, Table S4).

Figure 8.

Allosteric effects of vBCL1-4 on WIN55,212-2 and 2-AG concentration-response curves at the CB₂R. Functional cAMP assay of 10 μM vBCL1 (red circles), vBCL2 (green circles), vBCL3 (blue circles), and vBCL4 (violet circles) co-incubated with increasing concentrations of (A) WIN55,212-2 and (B) endogenous 2-AG in CHO-K1 cells stably expressing CB₂R. Calculated fold change of WIN55,212-2 and 2-AG EC₅₀ values for each vBCL is presented as table inset. Concentration-dependent allosteric modulation of WIN55,212-2 and 2-AG concentration-response curve in presence of 1 μM (inverted triangles), 3 μM (diamonds), and 10 μM (squares) of (C and D) vBCL3 (blue) and (E and F) vBCL4 (violet). Data are normalized to percentage of maximal activation, detected at the highest WIN55,212-2 and 2-AG concentration, and are shown as mean ± SD of 3 to 4 independent experiments. CB₂R, cannabinoid type 2 receptor; vBCL, vodo-C1-inspired bicyclic loop.

Given the interesting properties of vBCL3 (reduction of efficacy) and vBCL4 (greatest shift in potency), we further assessed the concentration-dependent effects of these two peptides when co-incubated with WIN55,212-2 and 2-AG (1 μM, 3 μM, and 10 μM). Similarly to their co-incubation with CP55940, vBCL3, and vBCL4 induced concentration-dependent rightward and downward shifts of the WIN55,212-2 and 2-AG concentration-response curves (8c-f, Tables S7 and S8). Operational model analysis was not possible for vBCL3, but vBCL4 was best fit using the operational model for allosterism, rather than Schild regression (slope <1), exhibiting an affinity of ∼646 nM when co-incubated with WIN55,212-2 (pK_B = 6.2 ± 0.4) and ∼300 nM when co-incubated with 2-AG (pK_B = 6.5 ± 0.4; Table 1). Functional cooperativity analysis (log_αβ) between CP55940,WIN55,212-2, and 2-AG for vBCL4 displayed ∼2- to 4-fold difference, indicating probe-dependent allosteric modulation of CB₂R with these three agonists, albeit this was not significantly different.

Lastly, we confirmed that the observed effects in cAMP assay were CB₂R-dependent, as no changes in cAMP levels were observed for vodo-C1 or CP55940 and 2-AG alone or in the presence of 10 μM of vBCL1-4 in untransfected CHO-K1 cells (Fig. S12, A–C).

Some allosteric modulators can influence ligand bias by modulating pathway-specific singling events (46). While vBCLs displayed inhibitory effects on G protein-mediated signaling pathways, we extended our study to pathway bias and explored their impact on the ability of CP55940 and 2-AG (both at 1 μM) to recruit β-arrestin-2. Co-treatment of the CB₂R with 10 μM vBCL2 and vBCL4 decreased the efficacy of CP55940 (p < 0.05) and 2-AG (n.s.) by ∼30%, whereas vBCL1 and vBCL3 exhibited no or only minor (not significant) changes in combination with CP55940 and 2-AG, respectively (Fig. 9, A and B).

Figure 9.

Allosteric effects of vBCL1-4 on 2-AG concentration-response curves at the CB₂R. BRET single-point measurements were conducted with 1 μM of (A) CP55940 and (B) endogenous 2-AG with or without 10 μM of vBCL1 (red circles), vBCL2 (green circles), vBCL3 (blue circles), and vBCL4 (violet circles) at HEK293 cells transiently expressing mouse CB₂R-eGFP and β-arrestin-2-Nluc. Statistical significance was tested using one-way ANOVA followed by Dunnett's post hoc test (∗p < 0.05). Concentration-response curves of (C) CP55940 and (D) 2-AG were measured in absence (black circles) and presence of 10 μM of vBCL4 (violet circles). Data are shown as the mean ± SD of 3 to 4 independent experiments. BRET, bioluminescence resonance energy transfer; CB₂R, cannabinoid type 2 receptor; Nluc, nano-luciferase.

Additionally, we demonstrated that 10 μM of vBCL4 decreased the potency (EC₅₀) of the agonists CP55940 and 2-AG by approximately 6-fold from 63 nM to 381 nM and 3-fold from 0.6 μM to 1.8 μM, respectively (Fig. 9, C and D, Table S9). These data suggest that vBCLs, in particular vBCL2 and 4 can weakly alter CB₂R-β-arrestin recruitment and do not display exclusive signaling bias for G protein-dependent signaling. However, more detailed concentration-dependent measurements in the future will strengthen these findings.

In summary, all four bicyclic peptides attenuated agonist-mediated cAMP inhibition of WIN55,212-2, 2-AG, and CP55940 albeit the effect was weaker in combination with CP55940. Furthermore, vBCLs displayed minor effects on the ability of CP55940 and 2-AG to recruit β-arrestin-2 to CB₂R. Detailed evaluation of the data suggests that vBCL4 is a negative allosteric modulator (NAM) on CP55940-, WIN55,212-2- and 2-AG-mediated inhibition of cAMP formation upon activation of CB₂R. Further, vBCL4 may be probe-dependent with the weakest effects observed for CP55940. Quantitative allosteric analysis was not possible for vBCL3 due to the nonsignificant effects on the EC₅₀-shift of CP55940 and WIN55,212-2 and the observed reduction in basal activity. Based on our data, vBCL2 (and vBCL1) may be best described as allosteric modulators of CB₂R, similar to vBCL4, whereas their lack of functional activation of CB₂R and partial displacement of radioligand in binding assays is also consistent with weak competitive (neutral) antagonism.

Discussion

Pharmacological targeting of CB₂R holds significant therapeutic potential for autoimmune diseases, (neuro-)inflammation, fibrosis, and chronic inflammatory pain (1, 2, 4, 47, 48). Our objective was to identify peptide ligands for this GPCR. Motivated by the chemical diversity and unique physicochemical properties inherent to nature-derived peptides as a reliable source of GPCR ligands, we systematically screened a custom library of cysteine-rich plant peptides to uncover modulators of CB₂R signaling (14, 19, 20, 49). Employing a robust pharmacology-guided fractionation approach led to the discovery of the peptide vodo-C1 from the sweet violet (V. odorata) as a novel CB₂R peptide agonist. The cyclotide-inspired design of bicyclic peptides vBCL1-4 based on the naturally occurring vodo-C1 sequence found in Viola spp. yielded novel probes with NAM properties at CB₂R signaling.

GPCRs stand out as one of the leading drug target classes (50). Typical approaches to identify GPCR-targeting peptide ligands involve high-throughput screening methods (51), in silico-genome mining (52), molecular grafting of stabilized peptide scaffolds (15, 53) and de novo-design (11). Since cannabinoid GPCRs are targeted by natural products including Δ⁹-tetrahydrocannabinol and its derivatives, fatty acid derivatives from Echinacea spp. (54), or the widespread plant sesquiterpene β-caryophyllene (55), among others, we utilized a peptide-enriched plant extract library screening approach. This led to the isolation and characterization of a peptide cannabinoid from sweet violet that showed a CB₂R full agonism.

Herbal preparations of sweet violet have been used in Persian and Indian ethnomedicine for centuries, particularly in treatment of pain, respiratory- and intestinal inflammation, and cancer (56, 57). In-depth investigations of V. odorata have acknowledged its expression of a variety of different cyclotides (35, 58). Still we identified and characterized a hitherto unknown cyclotide from this violet species leveraging peptidomics (24) and computer-assisted modeling (34). The larger volume of cyclotides, approximately eight times that of endogenous 2-AG and other cannabinoids, may account for the moderate affinity and potency of vodo-C1 at CB₂R. The bulkier size might impede deep penetration of the receptor binding-core (18) (∼415–447 Å³), ideally suited for binding of small molecule cannabinoid ligands (41). Peptide ligands of cannabinoid receptors like vodo-C1, including the recently discovered venom peptides that target CB₁R (59) may preferably interact with the extracellular regions of the binding pocket. While molecular details of peptide-ligand interaction with the CB₂R remain speculative, the identification of vodo-C1 as a plant-derived peptide CB₂R agonist expands the list known phytocannabinoids. The use of cyclopeptides offers opportunities for the development of novel CB₂R-selective peptide cannabinoid ligands. Moreover, peptide ligands like vodo-C1 and vBCL1-4 may inspire the development of probes to delineate the mode of action of peptide endocannabinoids (pepcans), in particular pepcan-12 (RVD-hemopressin), which is generated from hemoglobin 2 via the propeptide pepcan-23 (60, 61). Although the exact modes of action of pepcans (hemopressins) remain unclear, Bauer et al. have shown that pepcan-12 (RVD-hemopressin) can exert opposite allosteric effects at CB₁R and CB₂R, respectively (62). The allosteric binding sites in CB₂R have recently gained attention (8), particularly with endogenous signaling molecules like hemopressins exhibiting potential antinociceptive effects (7, 62). Building on the concept of miniaturization of cyclotide derivatives (14), the design of vodo-C1-inspired bicyclic peptides using a fragment-based approach yielded potent GPCR modulators with allosteric properties toward CB₂R signaling, albeit unexpected. In fact, similar effects have been described for the phytocannabinoids (−)-trans-Δ⁹-tetrahydrocannabinol and (−)-cannabidiol at CB₁R. These two compounds are naturally synthesized from the common precursor cannabigerolic acid and are chemical isomers. While cannabidiol is bicyclic, tetrahydrocannabinol contains an additional benzene ring (63). Despite the structural similarities, tetrahydrocannabinol and cannabidiol exert distinct neurochemical effects in the brain and their mechanisms of action on CB₁R differ markedly. While psychotropic tetrahydrocannabinol is a partial agonist of CB₁R, cannabidiol was identified as NAM of CB₁R, devoid of any psychotropic effects (64, 65). By similarity, the plant-derived cyclotide is an agonist of CB₂R, while the bicyclic peptide analogs are NAMs of CB₂R.

To date, only a limited number of CB₂R allosteric ligands has been identified, and the allosteric binding site has been postulated of being close to the N-terminal extracellular regions (66). The extracellular loop 2 of class A GPCRs plays a significant role in ligand binding and receptor activation, as it facilitates the access of the ligand to the orthosteric binding pocket, and interacts with allosteric modulators (67, 68, 69). Knowing that peptides are less capable of penetrating deeply into the hydrophobic binding pocket of CB₂R, as compared to small molecules (70), it is conceivable that vodo-C1 bicyclic analogs interact with extracellular loop 2 to manifest their mode of action.

The promising avenue of allosteric modulation in drug development has gained attention for its enhanced specificity and drug action (71, 72). Investigating the impact of vBCL peptides at CB₂R, we observed the reduction in potency and efficacy of synthetic and endogenous agonists. Interestingly, allosteric modulation by certain peptides is probe-dependent, highlighting the importance of selecting orthosteric probes for screening of allosteric modulators at CB₂R and other GPCRs in the future. Notably, allosteric modulators of CB₂R, like synthetic EC21a, have shown therapeutic potential, in several animal models (73, 74), indicative of the growing significance of allosteric modulation as an approach for targeting GPCRs. Specifically, it may offer an alternative strategy for enhancing cannabinoid receptor subtype selectivity of CB₂R. Negative modulation of CB₂R signaling could, for instance, hold therapeutic value for immunomodulation. Emerging evidence suggest that CB₂R plays a role in immune cell migration and wound healing, highlighting the possibility of using the CB₂R specific inverse agonists as immunomodulators (75). Recent studies have further supported a role of inverse CB₂R agonists in the regulation of microglial activation, as well as the attenuation of kidney injury in mouse models of renal fibrosis (76, 77). Therefore, fine-tuning of CB₂R signaling through application of negative allosteric modulators represents a compelling opportunity for future treatment approaches while minimizing side-effects of CB₂R agonists.

In summary, we have unveiled a novel phytocannabinoid peptide that acts as a full agonist at CB₂R, underscoring the untapped potential of nature, particularly of plant-derived peptides, as a valuable source for GPCR drug discovery. Intriguingly, the design of cyclotide-inspired bicyclic peptides led to an unexpected transformation of the molecular mode of action from an agonist to the discovery of modulators with negative allosteric and neutral antagonist properties for CB₂R. These peptides offer valuable tools for exploring cannabinoid receptor pharmacology and lay the foundation for future developments in allosteric peptide-based cannabinoid ligands, promising desired pharmacologic properties.

Experimental procedures

Materials

Radioligand [³H]-CP55940 was obtained from PerkinElmer. 2-Arachidonoylglycerol, AM630, CP55940, and (R)-(+)-WIN55,212-2 mesylate salt, ammonium hydrogen bicarbonate (NH₄HCO₃), polyethylenimine, tris(hydroxymethyl) aminomethane (Tris), magnesiumchloride hexahydrate (MgCl₂), forskolin, 3-Isobutyl-1-methylxanthine (IBMX), fatty acid-free bovine serum albumin (BSA), EDTA, α-cyano-hydroxy cinnamic acid (α-CHCA), and cell culture media and supplements were obtained from Sigma-Aldrich Chemicals. Acetonitrile (AcN), methanol (MeOH), dichloromethane, trifluoroacetic acid (TFA) were obtained as HPLC grade from Carl Roth. Endoprotease Glu-C, trypsin, and chymotrypsin were purchased from New England Biolabs. All other chemicals were of analytical grade and obtained from standard commercial sources. The jetPRIME transfection reagent was obtained from Polyplus and cAMP G_i kit from CisBio-PerkinElmer.

Cell culture, transfection, and cloning

CHO-K1 (American Type Culture Collection (ATCC), CCL-61) human CB₂R stable cell lines were cultured in Ham’s F12 culture medium supplemented with 10% fetal bovine serum, 50 U/ml penicillin/streptomycin, and 0.4 mg/ml geneticin (G418) and grown at 37 °C and 5% CO₂. The human CB₂R was N terminally cloned into the pEGFP-N1 vector using HindIII and BamHI restriction sites. Cell transfection was performed with HEK293T cells (ATCC, CRL-3216) and jetPRIME transfection reagent according to the manufacturer’s protocol.

Plant extraction

The extracts of B. alba, C. ipecacuanha, C. limon, H. annuus, P. tomentosa, S. alba, S. kombe, V. odorata and V. tricolor (Alfred Galke GmbH, Germany; P. tomentosa was collected in Costa Rica, at the tropical research station La Gamba) and peptide-enriched fractions thereof have been prepared as described previously (19). Briefly, 50 g of dried plant material was extracted with 1 L of MeOH/dichloromethane mixture (1:1, v/v) under continuous agitation at room temperature overnight. Plant material was removed by filtration and 0.5 volumes of water (ddH₂O) was added to the extract, to separate the aqueous MeOH phase containing peptides from organic phase. Prior to C₁₈ SPE the aqueous phase was evaporated and lyophilized, and the crude extract was dissolved in solvent A (99.9% ddH₂O, 0.1% TFA, v/v). The C₁₈ material ZEOprep 60 Å, irregular 40 to 64 μm (Zeochem) was equilibrated with solvent A and dissolved crude extract was loaded onto the C₁₈ cartridge. After washing with 10 to 30% of solvent B (90% AcN, 9.92% ddH₂O, and 0.08% TFA, v/v/v) the peptide-containing fractions were eluted with 50 to 80% solvent B, optimized for each plant extract.

Peptide analysis with MALDI and ESI mass spectrometry

MALDI-TOF MS analysis of peptide-enriched fractions and synthesized cyclotide was performed on MALDI-TOF/TOF 4800 Analyzer (AB Sciex). MS and MS/MS spectra were recorded in a reflector positive ion mode acquiring 2000 to 10,000 total shots per spectrum with a laser intensity of 3500. Samples were prepared by mixing 3 μl of α-CHCA and 0.5 μl of peptide and spotting 0.5 μl of the mixture onto the MALDI 384 target plate. Spectra were acquired, processed, and analyzed using the Data Explorer Software (AB Sciex). The molecular weights of synthetic bicyclic peptides were analyzed by electrospray ionization mass spectrometry (ESI-MS) using a LCMS 2010 system (Shimadzu).

RP-HPLC fractionation and peptide purification

The pharmacology-guided fractionation of the V. odorata peptide-enriched extract and purification of synthesized peptides was performed as previously described (19). In brief, the extract was dissolved in 5% solvent B and loaded onto the preparative Phenomenex Jupiter C₁₈ column (250 mm × 21.2 mm, 10 μm, 300 Å; Phenomenex). The mobile phase was composed of solvent A and solvent B. The automatic fractionation was performed on a Dionex 3000 LC machine (Dionex) using a linear gradient of solvent B between 5% and 65% at a flow rate of 8 ml/min. For analytical RP-HPLC runs, a Kromasil C₁₈ column (250 mm × 4.6 mm, 5 μm, 100 Å) and a Phenomenex Jupiter C18 column (150 × 2 mm, 5 μm, 300 Å) were used with applied linear gradient of solvent B between 5% and 65% and at a flow rate of 0.3 ml/min or 1 ml/min, respectively. The elution of peptides was monitored via UV absorbance at 214, 254, and 280 nm wavelengths.

Reduction, alkylation, and proteolytic digestion for de novo sequencing

Disulfide bond reduction of purified peptide was carried out by adding DTT (final concentration 10 mM, pH 8.5) to an aliquot of peptide sample (5 μg) previously dissolved in 0.1 M NH₄HCO₃ followed by incubation for 3 h at 60 °C. Cysteine residues were carbamidomethylated by adding iodoacetamide (final concentration 50 mM) to the reduced peptide sample and incubated for 1 min at 65 °C. Prior enzymatic digestion, the reaction was terminated by adding DTT (final concentration 10 mM) and incubated for 10 min at room temperature. Reduced and alkylated peptide was proteolytically cleaved by adding either EndoGluC (0.5 μg) or trypsin (0.4 μg) and chymotrypsin (0.4 μg) and were incubated for 3 h at 37 °C. The reaction was quenched by adding TFA (final concentration: 3%) to the samples. The masses of reduced, alkylated, and digested peptides were monitored by MALDI mass spectrometry. The identified corresponding precursor masses derived from peptide digestion were used for MS/MS fragmentation experiments. Acquired MS spectra were analyzed, and peptides were manually sequenced by assembling fragments of identified N-terminal b-ions and C-terminal y-ions series. The identified amino acid sequence was confirmed by High-Sensitivity Amino Acid Analysis (Macquarie University, Sydney, Australia).

Synthesis of vodo-C1 and bicyclic peptides

Cyclotide synthesis was carried out on a PTI Tribute Automated Peptide Synthesizer (PTI Instruments) utilizing Fmoc chemistry as previously described (78). 2-chlorotrityl chloride resin was loaded with hydrazine, followed by the coupling of the first amino acid. Further coupling reactions were carried out with amino acids, 2-(1H-benzotriazol-1-yl)-1,1,3,3-tetramethyluronium-hexafluorophosphat and diisopropylethylamine for 30 min as well as 2 × 5 min Fmoc deprotection with 20% piperidine in dimethylformamide. The full-length precursor was obtained after cleavage from the resin in a mixture of TFA/triisopropylsilane/ddH₂O/1,2-ethanedithiol (92.5%/2.5%/2.5%/2.5%; v/v/v/v) for 4 h at room temperature. The freeze-dried hydrazine peptide was dissolved in 0.2 mM phosphate buffer (pH 3) at the concentration of 2 mM and the cyclization was initiated with sodium nitrite at −20 °C for 15 min. The reaction was quenched with a 2-mercaptoethanesulfonate solution in phosphate buffer (pH 7.5) to a final peptide solution of 0.2 mM for 4 h. The cyclic reduced peptide was purified via SPE C₁₈ and freeze dried. To obtain the native cystine knot, peptide was dissolved in the folding buffer consisting of 0.1 M NH₄HCO₃ (pH 8.4)/isopropanol (1:1, v/v), supplemented with the oxidative shuffling reagents reduced (2 mM) and oxidized (0.5 mM) glutathione at a concentration of 0.5 mg/ml. To quench the reaction, the buffered sample was acidified to pH of ∼2 and freeze dried. The cyclic folded peptide was isolated with preparative HPLC, and the peptide purity was confirmed with analytical HPLC (214 nm) and MALDI. Vodo-C1 derived peptides were synthesized on a Microwave-Assisted Peptide Synthesizer (Liberty Prime, CEM) using a rink amide AM resin. Fmoc deprotection was done with 20% piperidine in dimethylformamide for 1 min at 90 °C. Amino acids were coupled in dicyclohexylcarbodiimide and ethyl cyanohydroxyiminoacetate (Oxyma), for 4 min at 90 °C. Cyclization of bicyclic peptides was done as described previously (42). Crude linear peptides were dissolved in buffer consisting of 70% (v/v) 20 mM NH₄HCO₃ (pH 8.0) and 30% (v/v) AcN and subjected for coupling reaction with 1,3,5-tris-(bromomethyl)-benzene for 1 h at room temperature. All peptides were purified using RP-HPLC, and peptide purity was confirmed with analytical HPLC runs at 0.3 and 1 ml/min. Correct peptide masses were confirmed with MALDI-MS or ESI-MS.

Radioligand binding assays

Radioligand binding assays using [³H]-CP55940 were carried out as described previously (8, 55). Membranes were prepared using CHO-K1 cell line stably expressing human CB₂R according to previously published protocols (79). All experiments were performed in duplicates and binding buffer containing 50 mM Tris–HCl, 2.5 mM EDTA, 5 mM MgCl₂, and 0.5 mg/ml fatty acid-free BSA (pH 7.4), in silanized glass vials. For competition binding, 75 μl each of [³H]-CP55940 (0.3–0.5 nM final), ligands (4X), and membranes (1–2 μg or 25 μg for CB₂R and CB₁R, respectively) were incubated for 2 h at 30 °C. To measure two-point radioligand displacement by bicyclic peptides, we used mixture of radiolabeled (0.5 nM) and cold CP55940 (4.5 nM). Nonspecific binding of radioligand was determined in presence of AM630 or WIN55,212-2 at final concentration of 10 μM. After the incubation, membrane suspensions were rapidly filtered through a 0.1% polyethylenimine-presoaked GF/B glass fiber filters (Sartorius Stedim Biotech) with Skatron cell harvester (Skatron AS) and washed 3 times with ice-cold washing buffer containing 10 mM TRIS–HCl, 1 mM MgCl₂, and 1 mg/ml fatty acid-free BSA (pH 7.7). The radioactivity retained on the filters was measured by liquid scintillation.

cAMP assay

The quantification of cAMP levels was carried out in triplicates with CHO-K1 cells stably expressing human CB₂R using the homogenous time-resolved fluorescence resonance energy transfer cAMP-Gi kit according to the manufacturer’s instructions with minor modifications as described previously (80). Briefly, 5000 cells per 5 μl per well were seeded into white 384-well plate and incubated with stimulation buffer (Opti-MEM media supplemented with 2% BSA and IBMX at 0.5 mM final concentration). Test compounds (4X) were diluted in stimulation buffer supplemented with 10 μM forskolin to final solvent concentrations ranging from 0.03 to 30,000 nM and coincubated with cells for additional 30 min at 37 °C. To measure the allosteric modulation, the cells were pretreated with peptides (4X) for 30 min at 37 °C followed by coincubation of CP55940, WIN55,212-2 or 2-AG (4X), and forskolin for 30 min at 37 °C. After the addition of 5 μl europium cryptate-labeled cAMP and cAMP d2-labeled antibody each and an incubation for 1 h at room temperature, fluorescence was measured at 620/665 nm using a Flexstation 3 plate reader (Molecular Devices). All measurements were performed with CP55940, WIN55,212-2, and/or 2-AG concentration-response curves as controls to assure cellular performance.

Bioluminescence resonance energy transfer assay

The measurement of β-arrestin-2 recruitment was carried out using cells cotransfected with plasmids transiently expressing human β-arrestin-2 nano-luciferase (Nluc) and human CB₂R-eGFP in 1:10 ratio. After at least 16 h, transfected cells were plated in white clear-bottom cell culture plates in Ham’s F12 culture medium supplemented with 10% fetal bovine serum at a density of 100,000 cells in 100 μl per well and allowed to adhere overnight. On the next day, the cells were serum-starved for 1 h at 37 °C in phenol red-free Dulbecco's modified Eagle's medium supplemented with 1% BSA. Furimazine (Promega), diluted 1:50, and ligand concentrations were prepared (4x) in Hank’s balanced salt solution and in duplicates. Furimazine was added to the cells and incubated for 5 min at 37 °C followed by the addition of CB₂R agonists CP55940 or 2-AG with or without peptide ligands and their incubation at 37 °C for 5 min. Plates were read for both luminescence at 460 nm for nano luciferase and fluorescence at 510 nm for enhanced GFP using the Flexstation 3 (Molecular Devices).

Data analysis

Data analysis was performed using GraphPad Prism (https://www.graphpad.com/) and statistical analysis was performed by an unpaired t test or one-way ANOVA. K_i values obtained from radioligand competition binding assays were determined by fitting the data to a three-parameter logistic Hill equation and applying the Cheng and Prusoff approximation (81) by using a previously determined K_D of [³H]-CP55940 (0.5 nM) (8). Data were normalized to specific binding of [³H]-CP55940 in absence of compounds as maximum percentage (100%), which refers to an average of 4500 to 5000 fmoles/mg protein for CB₂R and 250 to 350 fmoles/mg for CB₁R, respectively. To obtain dose response curves of functional assays, data were fitted to three-parameter nonlinear regression curves with a bottom and top constrained to 0 and 100, respectively, and a slope of one to obtain potency (EC₅₀) and maximum efficacy (E_max). Concentration response curves of functional assays for measuring allosteric modulation were generated by fitting the data to three-parameter nonlinear regression curves without constraints and a slope of one except for CP55940 which were constrained to zero (bottom) and hundred (top). Graphs were normalized to 100% which corresponds to the highest concentration of the positive control, which is either CP55940, WIN55,212-2, or 2-AG used in the assay. Concentration-response curves for the interaction between CP55940 or WIN55,212-2 and varying concentrations of each of vBCL1-4 in the cAMP accumulation assay were globally fitted to the following simplified operational model of allosterism and agonism, Equation 1 (82),where E_m denotes the maximum system response (efficacy).

$$E=Basal+\frac{\left(E_m-Basal\right)\left(\left[A\right]\left(K_B+\alpha \beta \left[B\right]\right)+{\tau }_B\left[B\right]{EC}_{50}\right)}{{EC}_{50}\left(K_B+\left[B\right]\right)+\left(\left[A\right]\left(K_B+\alpha \beta \left[B\right]\right)+{\tau }_B\left[B\right]{EC}_{50}\right)}$$ Equation 1

[A] and [B] are the concentrations of orthosteric agonist and allosteric ligand, respectively. K_B is the equilibrium dissociation constants of the allosteric ligand (vBCL1-4) and EC₅₀ is the half maximal response of the orthosteric ligand. τ_B is the operational efficacy of the allosteric ligand. αβ represents the composite of both binding and efficacy cooperativity factors between the orthosteric and allosteric ligands. Where K_B and αβ values were estimated, the analysis fixed the transducer slope (n) to 1, the Basal to 0 and E_m to 100.

Cyclic peptide modeling with AlphaFold

3D peptide structure predictions were performed as described previously (34). Briefly, AlphaFold (https://github.com/google-deepmind/alphafold) was used combined with modified workflow that enhances the accuracy and confidence when predicting and designing cyclic peptides. The advanced protocol termed AfCycDesign utilizes a cyclic offset matrix. N and C terminus were manually connected using the CONECT syntax in the Protein Data Bank file.

Molecular visualization and volume calculations

Peptide structures were visualized using PyMOL 2.5.5. Molecular volumes of compounds with a known 3D structure were calculated in UCSF ChimeraX 1.7 utilizing the built-in volume and area measuring tool. The volumes of molecules lacking 3D structure were calculated via Molinspiration’s interactive property calculator (https://www.molinspiration.com/).

Data availability

All data supporting this study are reported within this manuscript. Raw data are available from the corresponding author upon reasonable request. The cryo-EM structural data of CB₂R has been accessed via the Protein Data Bank (PDB) under accession code 6PT0, and the NMR structure of kalata B1 has been accessed via PDB accession code 1NB1. The sequence of vodo-C1 has been deposited to CyBase (http://www.cybase.org.au/). The structural model of vodo-C1 has been provided via GitHub (https://github.com/sHasinger27/VODO-C1-3D-structures).

Supporting information

This article contains supporting information (31, 34).

Conflict of interest

The authors declare that they have no conflicts of interest with the contents of this article.

Reviewed by members of the JBC Editorial Board. Edited by Kirill Martemyanov