Isolation of Positive Modulator of Glucagon-like Peptide-1 Signaling from Trigonella foenum-graecum (Fenugreek) Seed*

Background: Fenugreek is an edible plant used to treat diabetes and other disorders. Results: We isolated a new structure in fenugreek that enhances glucagon-like peptide-1 (GLP-1) potency. Conclusion: The antidiabetic plant fenugreek is a rich source of GLP-1 activity-enhancing compounds. Significance: GLP-1 peptide is a target for new drug screening and drug safety studies. The glucagon-like peptide-1 receptor (GLP-1R) is expressed in many tissues and has been implicated in diverse physiological functions, such as energy homeostasis and cognition. GLP-1 analogs are approved for treatment of type 2 diabetes and are undergoing clinical trials for other disorders, including neurodegenerative diseases. GLP-1 analog therapies maintain chronically high plasma levels of the analog and can lead to loss of spatiotemporal control of GLP-1R activation. To avoid adverse effects associated with current therapies, we characterized positive modulators of GLP-1R signaling. We screened extracts from edible plants using an intracellular cAMP biosensor and GLP-1R endocytosis assays. Ethanol extracts from fenugreek seeds enhanced GLP-1 signaling. These seeds have previously been found to reduce glucose and glycated hemoglobin levels in humans. An active compound (N55) with a new N-linoleoyl-2-amino-γ-butyrolactone structure was purified from fenugreek seeds. N55 promoted GLP-1-dependent cAMP production and GLP-1R endocytosis in a dose-dependent and saturable manner. N55 specifically enhanced GLP-1 potency more than 40-fold, but not that of exendin 4, to stimulate cAMP production. In contrast to the current allosteric modulators that bind to GLP-1R, N55 binds to GLP-1 peptide and facilitates trypsin-mediated GLP-1 inactivation. These findings identify a new class of modulators of GLP-1R signaling and suggest that GLP-1 might be a viable target for drug discovery. Our results also highlight a feasible approach for screening bioactive activity of plant extracts.

lungs, heart, kidneys, blood vessels, neurons, and lymphocytes (1). Mice with GLP-1R gene knock-out or with functionally deficient GLP-1R signaling exhibit impaired glucose homeostasis, learning, and memory (1). Clinical trials have also targeted GLP-1 signaling in other conditions, including psoriasis, heart disease, and neurodegenerative disease (2)(3)(4)(5). As such, GLP-1R signaling might be an ideal research target for the discovery and development of drugs for the treatment of various conditions from metabolic disorders to neurodegeneration. The GLP-1R is a current therapeutic target for treatment of type 2 diabetes (6,7).
After food intake, plasma levels of GLP-1 rapidly rise 3-4fold, enhancing glucose-dependent insulin secretion, to maintain normoglycemia (8,9). Bioactive forms of GLP-1 contain an alanine at position 2 and are rapidly degraded by dipeptidyl peptidase 4 to its inactive forms, GLP-1(9 -36)-amide or GLP-1(9 -37), following release from gut L cells. In general, GLP-1 is quickly secreted after food intake and rapidly cleared by dipeptidyl peptidase 4 and the kidneys (10). GLP-1 has a very short half-life and is sensed by its receptor in a transient and tightly regulated manner in healthy subjects.
Current therapeutic strategies aim to enhance activation of GLP-1R in target tissues by maintaining chronically high plasma levels of GLP-1 analogs (6,7). This approach is in contrast to the natural rapid secretion and clearance of GLP-1, and it can lead to loss of temporal regulation of GLP-1R signaling. GLP-1R is present in many different tissues and is involved in diverse physiological functions (1). It is often the case that local concentrations of GLP-1, rather than systemic levels, regulate tissue-specific GLP-1R activation (11). Therefore, constitutive activation of GLP-1R could result in various off-target effects; chronically high plasma GLP-1 levels can also lead to loss of spatial or tissue-specific control of GLP-1R activation. In some conditions, tissue-specific impairment of GLP-1R signaling, rather than a global decrease in plasma GLP-1, leads to disease pathogenesis, as is the case for type 2 diabetes (12)(13)(14). These findings indicate that constitutive activation of GLP-1R by chronically high plasma levels of GLP-1 analogs are likely not * This work was supported by Academia Sinica, Taiwan. R.-J. C. and K. K. have filed a patent application covering chemical compositions and applications for developing N55 analogs. the best therapeutic strategy. Clinical observations also suggest that side effects, such as nausea, vomiting, and gastrointestinal distress, may be related to chronically high plasma GLP-1 (15). Concerns about risks for developing pancreatitis and pancreatic malignancies with therapies targeting GLP-1R signaling have also been raised (16). These findings emphasize that other strategies should be considered when treating GLP-1R-related disorders.
To avoid these potentially adverse effects of treatment with GLP-1 analogs, we searched for positive modulators of GLP-1R signaling that themselves do not activate GLP-1R. Compounds that positively modulate GLP-1/GLP-1R signaling would control the degree of activation according to endogenous levels of GLP-1 and are less likely to lead to chronic activation of GLP-1R. Furthermore, such compounds would not affect temporal control of receptor activation.
To this end, we utilized two independent assay platforms designed to detect positive modulators of GLP-1R activity from crude plant extracts. One assay relates to GLP-1R internalization, a common step for activation of most of the G proteincoupled receptors (GPCRs) (17). Because activation of GLP-1R leads to an increase in intracellular cAMP production (18), the second assay employs a real time intracellular cAMP biosensor (RG-cAMP sensor) (19), capable of sensing intracellular cAMP production within minutes. These two assays were used in our initial screening of GLP-1 modulatory activity and for tracking and quantifying such activity during compound purification. In screening extracts from edible plants, we identified fenugreek seed extract that potentiates both GLP-1-dependent GLP-1R endocytosis and cAMP response. Fenugreek seeds have traditionally been used in Middle Eastern, Indian, Mediterranean, and Central Asian cuisine. Fenugreek seeds may also help with reducing blood sugar levels in diabetic patients (20). In addition, fenugreek seed also displays neuroprotective effects (21) and anti-inflammatory properties in animal models (22,23), these effects may be related to GLP-1R signaling.
The ability to potentiate GLP-1 potency would be a highly desirable characteristic of small molecules for the purpose of targeting GLP-1R signaling pathways. The methods we describe here represent important conceptual and operational approaches in discovering class B1 positive modulators of GPCRs. We report the isolation of an active compound with a new structure (N55) from fenugreek seeds. N55 itself does not trigger GLP-1R signaling but instead binds to GLP-1 and enhances its potency in stimulating GLP-1R.
cAMP Assay-Real time intracellular cAMP assay was performed as described previously (19). In brief, RINm5F cells stably expressing the RG-cAMP protein were seeded at 3 ϫ 10 4 cells/well in 96-well white plates in 0.15 ml of RPMI 1640 medium containing 400 g/ml G418. The next day, cells were washed twice with 0.1 ml of phenol red-free MEM containing 5 mM HEPES and incubated in the same medium for 1 h. The medium was replaced with 90 l of the same medium containing 1 mg/ml BSA and 5 M DeepBlueC. The whole plate was immediately loaded onto a SpectraMax Paradigm Detection Platform equipped with a Dual-Color luminescence detection cartridge and SoftMax Pro 6.2.2 (Molecular Devices, Sunnyvale, CA) to obtain the background BRET signal based on the sequential integration of the luminescence detected at 370 -450 and 500 -530 nm over 60 -150 s. Each well was then stimulated by adding 10 l of 10ϫ solutions of peptide and lipids and 1 mg/ml BSA in phenol red-free MEM containing 5 mM HEPES, and BRET signals were obtained immediately under identical settings. The BRET ratio is the ratio of light emitted between 90 and 300 s at 500 -530 nm to that emitted at 370 -450 nm. The cAMP response was expressed as a percentage of cAMP production and was calculated as 100 ϫ (BRET ratio from 0.01 nM GLP-1(7-36)-amide Ϫ BRET ratio from indicated concentration of peptide with or without lipids)/(BRET ratio from 0.01 nM GLP-1(7-36)-amide Ϫ BRET ratio from 250 nM GLP-1(7-36)-amide). The dose-response curve, maximal response, and concentration of peptide needed to yield halfmaximal response (EC 50 ) were obtained by nonlinear regression to fit the data to the agonist versus response equation using Prism software 5.0 (GraphPad, San Diego). Unless specified, all cAMP response data are the means Ϯ S.E. from three independent experiments with triplicate assays.
Receptor Endocytosis Assay-U2OS osteosarcoma cell line stably expressing a ␤-arrestin2:GFP fusion protein was obtained from Norak Biosciences. The pcDNA3 GLP-1R-V2R chimeric construct contains the first 440 amino acids of the GLP-1R (Met-1 to Thr-440) fused to the last 29 amino acids of the vasopressin V2 receptor (Ala-343 to Ser-371) (24) and separated by two alanine residues as linker. GLP-1R-V2R chimeric construct is inserted into the EcoRI site of pcDNA3 (pcDNA3-GLP-1R-V2R) such that expression of the chimeric protein is under the control of the CMV promoter. pcDNA3-GLP-1R-V2R was used to transfect U2OS osteosarcoma cells stably expressing ␤-arrestin2:GFP to obtain a cell line stably co-expressing GLP-1R-V2R and ␤-arrestin2:GFP. High content imaging of receptor endocytosis in cells was conducted with 0.03 to 0.001 mg/ml extract to identify potentiating activity for GLP-1-dependent GLP-1R endocytosis. Extract were supplied at a concentration of 100 mg/ml in 100% DMSO. Three replicate 384-well assay microplates were plated with U2OS cells stably co-expressing GLP-1R-V2R and ␤-arrestin2: GFP at a density of 3 ϫ 10 3 cells/well. Aliquots of 2.5 l of 10ϫ stocks of the indicated concentration of extract in phenol redfree MEM containing increasing concentrations (0.12-3000 nM) of GLP-1(7-36)-amide and 1 mg/ml BSA were transferred to each well of the assay plate, which contained 22.5 l of phenol red-free MEM containing 1 mg/ml BSA. The three assay plates were incubated at room temperature for 60 min before cell fixation with 2% formaldehyde and labeling of the cell nuclei with 5 g/ml of the DNA-binding dye Hoechst 33342 for 1 h. Plates were washed twice with PBS and sealed; plates were used immediately.
Imaging and Analysis-Images were acquired on an XL model of the ImageXpress Micro System (Molecular Devices, Sunnyvale, CA) and analyzed with MetaXpress High-Content Image Acquisition and Analysis Software (Molecular Devices, Sunnyvale, CA) using the granularity and variable grain analysis modules. MetaXpress was used to retrieve images using DAPI (to retrieve the blue fluorescent Hoechst 33342-labeled nuclear images) and FITC (to retrieve the green fluorescent GFP-␤arrestin images) filter sets and a Plan Fluor ELWD objective. A 20 ϫ 0.45-numerical aperture microscope objective was used for the imaging; two fields were imaged per well. The number of spots, total area covered by spots, average and integrated fluorescence intensity of the spots, and nuclear area and fluorescence intensity were logged into the database. Data from abnormal cells for which the values were above threshold (abnormal: average nuclear staining intensity Ͻ500; integrated fluorescence intensity of area covered by spots Ͻ150) were ruled out using Acuity Xpress, followed by selection of the measurement method, spot fluorescence integrated intensity to calculate the average number of spots per nucleus. Dose-response curve, maximal response, and the concentration needed to yield half-maximal response were obtained using nonlinear regression to fit the data to the agonist versus response equation with Prism software 5.0. The extent of receptor endocytosis response was expressed as percent of that elicited by 750 nM GLP-1(7-36)-amide.
Competition Binding Assay-Competition assays to evaluate the binding of signal non-enhancing SEA and N55 to His-tagged GLP-1 were carried out as described previously (19). Briefly, increasing concentrations of N55 or SEA were incubated with 0.2 M [ 3 H]OEA in the presence (total binding) or absence (nonspecific binding) of 0.2 M His-tagged GLP-1. The reaction was carried out in 50 l of Dulbecco's phosphate-buffered saline (DPBS) containing 0.02 mg/ml bovine serum albumin (BSA) at room temperature for 90 min. Copper-iminodiacetic acid-agarose (Cu 2ϩ -agarose), 3 l in 30 l of DPBS, was added to capture all His-tagged GLP-1 (up to 60 M peptide in an 80-l reaction), and the mixture was further incubated with rotation at room temperature for 30 min. The mixture was centrifuged at 20,600 ϫ g for 10 min at 4°C to precipitate Histagged GLP-1 trapped in the resin, and 20 l of supernatant containing the unbound free [  Preparation of GLP-1 Receptor Membrane-U2OS cells stably expressing GLP-1R-V2R were grown to 90% confluence (about 10 7 cell per 15 cm dish). The media were removed and washed twice with 30 ml of PBS, followed by adding 2.2 ml of ice-cold homogenization buffer (20 mM HEPES, 1 mM EDTA, 0.7% protease inhibitor mixture P8340 (Sigma) per dish. The cells were scraped immediately and centrifuged at 3000 ϫ g, 4°C, for 30 min. The pellet was resuspended with 5 ml of homogenization buffer and then homogenized at maximal speed on a Polytron 3000 for 10 s on ice with a 30-s interval rest, three times. The homogenized cells were centrifuged for 10 min at 4°C and 1000 ϫ g. The supernatants were transferred to a fresh transparent centrifuge tube and centrifuged for 60 min at 4°C and 55,000 ϫ g. The pellets were resuspended with resuspension buffer (20 mM HEPES, 1 mM MgCl 2 , 0.7% protease inhibitor) by passing through a 22-gauge needle one time, a 25-gauge needle three times, then a 26-gauge needle one time. The protein concentration of the membrane preparation was determined according to the Qubit fluorometer instructions (Life Technologies, Inc.). 4 -5 g of this membrane yielded greater than 5-fold signal/background with 125 I-labeled GLP-1 at 5 nM in a 0.22-ml assay volume.
Assay for Insulin Release from RINm5F Cells-For measuring insulin release, RINm5F cells were seeded at 60,000 cells/well in a 96-well plate in 0.15 ml of RPMI 1640 medium. The next day, cells were washed twice with 0.1 ml of phenol red-free MEM (11 mM glucose) containing 5 mM HEPES, pH 7.0, and 1 mg/ml BSA; cells were incubated at the same medium for 1 h. The medium was replaced with 0.05 ml of the same fresh medium containing the indicated compound and were incubated for 30 min at 37°C. Media were withdrawn, centrifuged, and assayed for insulin by rat insulin ELISA (Mercodia AB, Uppsala, Sweden). Each assay was performed in duplicate, and each was repeated three times.
Cell Viability Test-Cytotoxicity of H460 cells to various compounds was assessed using PrestoBlue reagent, which is modified by the reducing environment of the viable cell and turns from blue to red in color. In brief, 3000 H460 cells were seeded with 0.1 ml of RPMI 1640 medium in each well of a 96-well plate and incubated at 37°C for 24 h before compound treatment. After a 72-h compound treatment, PrestoBlue TM cell viability reagent (Invitrogen) was added to each well according to the manufacturer's protocol. The plates were further incubated in the dark at 37°C in 5% CO 2 for 10 min, and absorbance of all wells was read at both the 570-and 600-nm wavelengths using automatic enzyme-linked immunosorbent assay plate reader (Molecular Devices, Union City, CA). The raw data were normalized and corrected according to the manufacturer's protocol. All samples were tested in triplicate, and each test was repeated three times. Cell viability was calculated using the following formula: percent of 100 ϫ (average corrected values of compound treated cells)/(average corrected values of untreated cells).
Receptor Binding Assay-Assays were conducted in 0.22 ml of 50 mM HEPES, pH 7.4, 5 mM MgCl 2 , 1 mM CaCl 2 , and 1 mg/ml BSA containing 3.9 g of GLP-1 receptor membranes and varying concentrations of 125 I-labeled GLP-1(7-36) radioactive ligand with or without 7.7 M N55 and in the absence (total binding) or presence (nonspecific binding) of 1000-fold excess unlabeled exendin 4. The reactions were incubated for 90 min. Prior to filtration, an FC 96-well harvest plate (Millipore, MAFC N0B 10; Billerica, MA) was coated with 0.5% polyethyleneimine for 30 min, then washed with 50 mM HEPES, pH 7.4. 0.1 ml of binding reaction was transferred to the filter plate and washed 15 times (0.1 ml per well per wash) with ice-cold 25 mM HEPES, pH 7.4, and 50 mM NaCl. The plate was dried followed by adding 30 l of MicroScint 20 (PerkinElmer Life Sciences) per well and, the activity was determined by using singlephoton counting (60 s/well read) on a TopCount scintillation counter (PerkinElmer Life Sciences). Specific bindings were obtained by subtracting nonspecific binding from total binding. Dissociation constant (K d ) for GLP-1(7-36)-amide was obtained using Prism software 5.0 (Graph Pad, La Jolla, CA). Data shown are the means Ϯ S.E. and were duplicated from three independent experiments.
In Vitro Assay of Trypsin Activity in the Presence of N55-To determine whether N55 affects trypsin activity, we used the trypsin activity colorimetric test kit (BioVision, Milpitas, CA). Briefly, 2 l of substrate were added to a well in a 96-well plate containing 48 l of 0.000125% trypsin in the presence or absence of 77 M N55. Reactions were conducted at room temperature, and the extent of cleavage was monitored by real time reading of absorbance at 405 nm.
Limited Trypsin Digestion of GLP-1(7-36)-Amide-Trypsin digestion was carried out at 37°C for 30 min in 0.1 ml of phenol red-free MEM containing 5 mM HEPES, pH 7.0, 2 M GLP-1(7-36)-amide, 2 ϫ 10 Ϫ3 to 2.47 ϫ 10 Ϫ5 % of trypsin (prepared by diluting 0.05% trypsin-EDTA), and the indicated concentrations of N55 or SEA. Reaction was terminated by incubating at 94°C for 30 min and then at 4°C for 10 min, followed by addition of PMSF to 1 mM, and 30 min of incubation at room temperature. N55 was adjusted to 7.7 M for all the cAMP responses to titration of the GLP-1 in RINm5F cells.
Isolation of N55 from Fenugreek Seeds-Ethanol extraction of fenugreek seeds and solvent partition of the extracts are outlined in Fig. 1. Ethyl acetate fraction of solvent partition (FE-E-EA) was further fractionated by normal phase silica gel chromatography. 25 g of FE-E-EA was absorbed onto 79 g of silica gel and then chromatographed on a column (4.5 ϫ 43 cm) packed with 370 g of silica gel by stepwise elution with 2500 ml of 20:80 ethyl acetate/hexane (20% EA/Hx), 2500 ml of 40:60 ethyl acetate/hexane (40% EA/Hx), 2500 ml of 60:40 ethyl acetate/hexane (60% EA/Hx), 2500 ml of 80:20 ethyl acetate/ hexane (80% EA/Hx), and 2500 ml of 100% ethyl acetate (100% EA), followed by 2500 ml of methanol/ethyl acetate (50:50; 50% EA/Me). Thirty 500-ml fractions were collected, evaporated, and subjected to activity tests, and fractions with expected activities were combined. Final purification of the combined F2 fraction and primary structure elucidation of the active compound was contracted to AnalytiCon Discovery GmbH (Potsdam, Germany). Briefly, 3 g of the partially purified extract with expected activity from normal phase silica gel chromatography (F2) was further fractionated by preparative reverse phase-HPLC (RP-HPLC) (LiChrospher select B column 250 ϫ 25 mm, 10 m). The column was first eluted using a linear gradient of 50 -95% acetonitrile/methanol in water for 57.7 min at a flow rate of 80 ml/min, followed by elution with 100% acetonitrile/methanol for 10 min. Based on detected peaks (evaporative light scattering detector and UV light), the neighboring similar fractions per fractionation were partially combined to give 72 fractions. The resulting final fractions were evaporated and subjected to activity tests. Fractions exhibiting potent activity enhancing GLP-1R signaling through GLP-1 were subjected to structure elucidation.
Synthesis and Final Structure Characterization of N55-See supplemental material for synthesis and final structure characterization of N55.

Results
Fenugreek Seed Extracts Potentiate GLP-1 Signaling-Initial extraction and solvent partition of fenugreek seeds is illustrated in Fig. 1. To investigate GLP-1R signaling, we employed ␤-ar-restin2-GFP biosensing technology, based on the observation that ␤-arrestin binding of an activated receptor is a convergent step in GPCR signaling (17). These processes can be visualized (25), allowing for exclusion of false-positive hits. Because GLP-1R activation also leads to an increase in cAMP production (19), the RG-cAMP sensor (an Epac1 cAMP biosensor) (19) was used to monitor real time cAMP production in rat insulinoma RINm5F cells (19). This rapid real time cAMP response assay significantly reduced detection of nonspecific effects of the extract mixture and allowed us to detect receptorspecific responses.
Ethanol extracts of seeds of fenugreek (FE-E) enhanced the response to GLP-1(7-36)-amide stimulation. ␤-Arrestin2-GFP fusion proteins were originally evenly distributed in the cytosol of U2OS cells co-expressing ␤-arrestin2-GFP and GLP-1R ( Fig.  2A). The formation of pit complex ␤-arrestin2-GFP-GLP-1R can be visualized when cells were stimulated with 1 nM GLP-1(7-36)-amide. Pit complexes became more prominent when 0.1 mg/ml FE-E was included. FE-E alone did not induce formation of pits ( Fig. 2A). Fig. 2B illustrates the effect of FE-E on receptor endocytosis with titration of GLP-1(7-36)-amide. EC 50 of GLP-1(7-36)-amide required to induce half-maximal response was reduced from 3.8 to 1.7 nM by 0.1 mg/ml FE-E. This effect of FE-E is highly dependent on the concentration of GLP-1, as it was diminished when GLP-1 concentration was reduced to 0.01 nM (Fig. 2B). FE-E facilitated receptor endocytosis elicited by 1 nM GLP-1(7-36)-amide in a dose-dependent and saturable manner (Fig. 2C). Receptor endocytosis elicited by 1 nM GLP-1(7-36)-amide was no more than 10% of that elicited by 750 nM GLP-1(7-36)-amide. Addition of 0.0037 mg/ml FE-E enhanced receptor endocytosis by up to 20%, and the response was saturated at 33%, when the concentration of FE-E reached 0.03 mg/ml (Fig. 2C). EC 50 of FE-E required to enhance stimulation by 1 nM GLP-1 was 0.0035 mg/ml. FE-E was further subjected to fractionation by solvent partition to produce hexane (FE-E-Hx), ethyl acetate (FE-E-EA), and butanol fractions (FE-E-Bu) (Fig. 1), and their effects on receptor endocytosis were examined (Fig. 2, D-F). Signal-enhancing activity was enriched in FE-E-EA. EC 50 of extract required to enhance receptor endocytosis by 1 nM GLP-1 was reduced from 0.0035 mg/ml (FE-E) to 0.0009 mg/ml (FE-E-EA) ( Table 1). Because GLP-1R activation leads to intracellular cAMP production in pancreatic ␤-cells, we tested whether these extracts potentiate GLP-1-induced cAMP production in RINm5F cells. We examined the effect of 0.1 mg/ml FE-E-Hx and FE-E-EA fractions on the potency of GLP-1 stimulation of cAMP production (Fig. 3). GLP-1 stimulated cAMP production in RINm5F cells expressing the RG-cAMP sensor (19) in a dosedependent and saturable manner, with an EC 50 of 2.8 nM (Fig.  3). In the presence of 0.1 mg/ml FE-E-EA or FE-E-Hx, EC 50 was reduced to 0.18 or 1.67 nM, respectively. The butanol and the water fractions did not show any activity. A summary of this step of fractionation is shown in Table 1. Activity of FE-E was highly enriched in the ethyl acetate fraction (FE-E-EA), whereas butanol and water fractions displayed little activity.
Isolation and Structural Elucidation of the Active Compound-Because the FE-E-EA fraction potently enhanced GLP-1R signaling, this fraction (24 g) was chromatographed on normal phase silica gel. The column was eluted with mixtures of proportional hexane, ethyl acetate, and methanol of increasing polarity. As shown in Fig. 4, the FE-E-EA fraction was separated into 30 fractions, which can be divided into four major groups according to their ability to potentiate GLP-1-induced GLP-1R endocytosis (Fig. 4A). The first group (F1) includes fractions 4 and 5; the second group (F2) includes fractions 7-12; the third group (F3) includes fractions 18 -22; and the fourth group (F4) includes fractions 24 -27. In addition to facilitation of GLP-1R endocytosis, we also examined the effect of these fractions (10 g/ml) on cAMP production induced by 1 nM GLP-1. Only fractions 8 -13 significantly potentiated cAMP response (Fig. 4B).
N55 Enhances GLP-1R Signaling-We analyzed the effect of N55 on GLP-1-induced cAMP production in RINm5F cells stably expressing RG-cAMP (19). GLP-1 induced cAMP production in a dose-dependent and saturable manner (Fig. 7A), with an EC 50 of 2.4 nM. N55 enhanced the cAMP response to GLP-1 and shifted the dose-response curve to the left Ͼ10-fold. N55 potently reduced the EC 50 of GLP-1 by a factor of 40 as its concentration increased from 0.8 to 26 M (Fig. 7A). The augmentation diminished as GLP-1 concentration fell to 1 pM, indicating a lack of intrinsic agonistic activity of N55 (Fig. 7A). N55 increased the potency of GLP-1 in a dose-dependent and saturable manner (Fig. 7B), with an EC 50

Fenugreek Compound Binds to and Enhances GLP-1 Potency
OCTOBER 23, 2015 • VOLUME 290 • NUMBER 43   GLP-1 was enhanced Ͼ5-fold by 7.7 M N55 (Fig. 7C, Student's t test, p Ͻ 0.001), and both cAMP response and enhancement were eliminated by 2 M of the GLP-1R antagonist Ex-9. This analysis indicates that the enhancing effect of N55 on GLP-1-induced cAMP response was solely mediated by GLP-1R in RINm5F cells. To examine the specificity of N55 on GLP-1R signaling, we further analyzed the effect of N55 on cAMP production in RINm5F cells stimulated by the gastric inhibitor polypeptide (GIP). GIP is another incretin, and its receptor is also expressed in RINm5F cells (27). Activation of the GIP receptor leads to production of cAMP in RINm5F cells (19,27). The cAMP response to either GIP or exendin 4 (Ex-4, a GLP-1 analog) is not affected by N55 (Fig. 7, E and D, respectively). This analysis indicates that the enhancing effect of N55 on cAMP production is specific to the GLP-1(7-36)-amide.

N55 Facilitates GLP-1-elicited Insulin Release from RINm5F
Cells-One major physiological response to the activation of the GLP-1 receptor in pancreatic ␤-cell lines is insulin release (1). As N55 enhanced the potency of GLP-1(7-36)-amide, we further examined whether N55 could enhance GLP-1-elicited insulin release from RINm5F cell, a rat insulinoma cell line that has been show to release insulin in response to GLP-1 stimulation (28). As shown in Fig. 9, 2 nM GLP-1(7-36)-amide consistently stimulated a low level of insulin release from RINm5F cells, and this response is significantly enhanced by 7.7 M N55, although N55 alone does not trigger insulin release. Intracellular cAMP level is primarily responsible for enhanced glucosedependent insulin secretion from pancreatic ␤-cells (29). Increasing intracellular cAMP level by forskolin will lead to insulin release from RINm5F cells (Fig. 9). This is consistent with the enhancement effect of N55 on cAMP production results from GLP-1 stimulating RINm5F cells.
N55 Does Not Affect Cell Viability-N55 alone does not affect receptor endocytosis, cAMP production, and insulin secretion. To examine whether these negative effects could result from the cytotoxicity of N55, we compared the effect of N55 on cell viability to those of endocannabinoid-like lipids (2-oleoylglycerol and oleoylethanolamide) and cisplatin (an anti-apoptosis compound). Within the test concentration from 0.78 to 100 M, N55, 2-oleoylglycerol, and oleoylethanolamide do not show detectable cytotoxicity; however, cisplatin potently killed the cell at a concentration of 50 M and displayed an IC 50 of 6.7 M (Fig. 10). This analysis revealed that N55 displayed comparable effects on cell viability as those of endocannabinoid-like lipids.
Effect of N55 on Saturation Binding of GLP-1 to GLP-1R-Because N55 enhanced the potency of GLP-1(7-36)-amide, we further investigated whether this effect may have been due to enhanced binding affinity of GLP-1(7-36)-amide to GLP-1R. We performed saturation binding assays to examine the effects of N55 on GLP-1R binding affinity. Binding of 125 I-labeled GLP-1(7-36)-amide to GLP-1R is dose-dependent and saturable, the dissociation constant (K d ) and maximal binding (B max ) being 1.5 Ϯ 0.25 nM and 51 Ϯ 2.9 pmol/mg membrane, respectively (Fig. 11A). N55 at a concentration of 7.7 M enhanced cAMP response 30-fold, although this concentration of N55 barely affected the saturation binding curve (Fig. 11B). K d and B max values in the presence of 7.7 M N55 was 1.4 Ϯ 0.5 nM and 51 Ϯ 6.5 pmol/mg of membrane, respectively. Specific binding of 125 I-labeled GLP-1(7-36)-amide to GLP-1R in the presence or absence of 7.7 M N55 is almost superimposed (Fig. 11C). This analysis suggests that N55 minimally affected GLP-1's affinity for and capacity to bind GLP-1R. Though enhancing the potency of GLP-1(7-36)-amide in stimulating the cAMP response, 7.7 M N55 did not affect the physical affinity between GLP-1 and its cognate receptor.
N55 Binds to GLP-1(7-36)-Amide and Facilitates Trypsin Inactivation of GLP-1(7-36)-Amide-N55 selectively enhanced the potency of GLP-1(7-36)-amide, but not that of Ex-4, possibly because N55 might interact with the GLP-1(7-36)-amide peptide. It has been shown that GLP-1 binds to signal-enhancing OEA or 2-oleoylglycerol but not to SEA (19). Competition binding experiments were performed to examine the effect of increasing N55 concentrations on the binding of [ 3 H]OEA to GLP-1. As shown in Fig. 12A, binding of OEA gradually decreased as the concentration of N55 increased from 0.025 to 1.5 M, and the inhibition constant (K i ) was estimated to be 0.03 Ϯ 0.01 M. In contrast, SEA did not affect the binding of OEA to GLP-1. A potential consequence of the binding of N55 to GLP-1 is induction of a conformational change in the peptide. It has previously been demonstrated that there are sig- nificant concentration-dependent conformational changes in GLP-1 within the range of 3-50 M (30). Therefore, we utilized susceptibility to trypsin digestion to explore potential conformational changes in GLP-1(7-36)-amide at low micromolar concentrations upon binding N55. The two trypsin cleavage products of GLP-1(7-36)-amide correspond to an inactive GLP-1 (7-26) and a partially active GLP-1 (7-34) (31). Residual activity simulating cAMP production after trypsin treatment was used to determine susceptibility of the peptide to trypsin digestion. Increasing the concentration of trypsin from 2.47 ϫ 10 Ϫ5 to 2 ϫ 10 Ϫ3% reduced residual GLP-1 activity by a factor of 27 (Fig. 12B). When the trypsin cleavage reactions were carried out in the presence of 77 M N55 (Fig. 12C), the residual GLP-1 activity was dramatically reduced by a factor of Ͼ300 (Fig. 12C). We also compared the effect of SEA, which was shown to have no effect on the potency of GLP-1 in stimulating cAMP production. Fig. 12D shows that residual GLP-1 activity was reduced only by a factor of 8 in the presence of 92 M SEA. These results illustrate that trypsin digestion of GLP-1(7-36)-amide was facilitated by N55, but N55 did not affect innate enzymatic activity of trypsin (Fig. 13). To determine whether this effect of N55 is dose-dependent, inactivation of GLP-1(7-36)-amide by 6.7 ϫ 10 Ϫ4% trypsin was carried out with increasing concentrations of N55 from 0.8 to 77 M. Residual activity of GLP-1(7-

Fenugreek Compound Binds to and Enhances GLP-1 Potency
OCTOBER 23, 2015 • VOLUME 290 • NUMBER 43 36)-amide decreased by a factor of Ͼ30 (Fig. 6E) as the concentration of N55 increased from 0.8 to 77 M. The effect of N55 became saturated as it reached a concentration of 7.7 M (Fig.  12E). These analyses indicate that N55 dose-dependently and saturably increased susceptibility of GLP-1(7-36)-amide to trypsin inactivation (Fig. 12F), with an EC 50 of 2 M. These findings suggest that a conformational transition in the GLP-1(7-36)-amide peptide occurred upon binding N55.

Discussion
Fenugreek seeds have been used as a treatment for diabetes and have been found to lower blood glucose levels in human trials (20,(32)(33)(34). We report the isolation and characterization of a new structure from fenugreek seeds. N55 binds to GLP-1 and enhances the potency of GLP-1 in stimulating GLP-1R signaling. This finding is consistent with research targeting GLP-1R signaling in the treatment of diabetes. Using agonistinduced receptor endocytosis and real time measurement of intracellular cAMP production, we purified and isolated N55 from fenugreek seeds. It is noted that although some fractions from silica gel chromatography did not potentiate cAMP production, they were still able to enhance receptor endocytosis (Fig. 4, A and B). Because cAMP is essential and sufficient for pancreatic ␤-cell secretion of insulin (29), we combined fractions with enhancing effects on cAMP production for further purification. As shown in Fig. 5, most of the fractions at this step   of isolation enhanced GLP-1-induced cAMP production, although fraction N55 is of high recovery and significantly pure. The effect of N55 on GLP-1R signaling is specific to the GLP-1(7-36)-amide. It does not enhance signaling by the GLP-1 analog Ex-4 (Fig. 7D) or the other incretin GIP (Fig. 7E). N55 does not activate GLP-1R by itself, as its effect is highly dependent upon the presence of GLP-1(7-36)-amide (Fig. 7A). N55 enhanced the potency of GLP-1 stimulation of the cAMP response mediated by GLP-1R signaling, and this augmentation is eliminated when GLP-1R signaling in RINm5F cells is antagonized by Ex-9 (Fig. 7C).
There are two possible mechanisms that may explain the effect of N55 on GLP-1R signaling. N55 can either act upon GLP-1R or on GLP-1(7-36)-amide. At least three lines of evi-dence favor the interpretation that N55 interacts with GLP-1(7-36)-amide. First, N55 selectively enhanced the potency of GLP-1(7-36)-amide but not that of Ex-4 (Fig. 7, A and D). GLP-1 and exendin 4 share 50% sequence identity but have distinct secondary structure, order, and stability characteristics (30,35). Second, N55, but not SEA (19), binds GLP-1 in a dosedependent manner (Fig. 12A). Finally, N55 specifically enhanced susceptibility of the peptide to cleavage by trypsin (Fig. 12, C and E), although SEA (19) did not (Fig. 12D). These are consistent with previous findings that only specific signaling enhancing lipids bind GLP-1, while also enhancing its susceptibility to trypsin digestion (19). These observations are consistent to the model that GLP-1 can bind to specific lipids to form a complex with enhanced potency in stimulating GLP-1R signaling. Covalent linking of a lipid molecule to peptide hormones (lipidation) is commonly used as a chemical tool to optimize the pharmacokinetic properties of peptide drugs (36). Moreover, lipidation may stabilize a structure of the peptide hormone, leading to broadening its spectrum of activity (37). Most of the lipidations of GLP-1 have no discernible effects on GLP-1R signaling, with a few exceptions, which enhanced the potency of GLP-1 by 6 -7-fold (38). In contrast to covalent linking, our studies show that non-covalent binding of N55 to GLP-1 also stabilized a peptide structure displaying a 40-fold enhancement in potency to stimulate cAMP production.
The postulated mechanism by which N55 enhances GLP-1R signaling is different from that of allosteric modulators isolated in the past (39 -42). All previously isolated GLP-1R modulators exert their effects by binding to allosteric sites on the receptor and display ago-allosteric modulation. These allosteric modulators display both allosteric agonism and an ability to allosterically modulate the binding, specificity, and/or function of the GLP-1 ligand. This effect may be due to the allosteric modulator causing a redistribution of conformationally linked receptor states, which can include signaling species. In contrast, N55 had a pharmacological profile distinct from that of GLP-1 modulators identified in the past; N55 does not display agonism or altered receptor specificity in the absence of GLP-1 ligand.
Fenugreek has been used for the treatment of diabetes in many parts of the world, particularly in China, Egypt, India, and Middle Eastern countries (43,44). In humans, fenugreek seeds acutely reduce post-prandial glucose and insulin levels (32). Several longer term clinical trials have also shown that administration of fenugreek seeds reduces fasting and post-prandial glucose levels and decreases glycated hemoglobin (33). Soluble fiber (45,46), saponins (47), trigonelline (48), diosgenin (49), and 4-hydroxyisoleucine (50) from fenugreek seeds have been suggested to mediate its hypoglycemic effect (33). The exact molecular targets and mechanisms remain unknown; however, N55 from fenugreek may potentially play a role in ameliorating glucose excursion in diabetes. In addition to hypoglycemic effect, extract of fenugreek seed also displays neuroprotective properties (21) and anti-inflammatory effects (22,23) in disease animal models. These mechanisms of effects are consistent with the current trials of GLP-1 analogs in treatment of psoriasis (5) and Parkinson disease (2).
Current GLP-1 analog therapies maintain constant high levels of GLP-1 activity, leading to constitutive activation of  GLP-1R and loss of spatiotemporal regulation of GLP-1R activity. Positive modulation of GLP-1R signaling may offer the opportunity to augment and/or fine-tune endogenous GLP-1 action. Enhancers of GLP-1 potency modulate GLP-1R signaling according to endogenous levels of GLP-1. They do not themselves activate the receptor, but serve only to augment the potency of circulating GLP-1. This property may be of particular importance for brain GLP-1R-expressing neurons, where activation of the receptor is regulated by local GLP-1-secreting fibers.
Deterioration in GLP-1 signaling has been associated with type 2 diabetes (13). Such patients have impaired incretin function, despite having normal or only mildly reduced GLP-1 secretion, as measured by oral glucose or meal tests (13,14). Clinically, administration of a GLP-1 analog has been successful in ameliorating the effects of reduced GLP-1 signaling in type 2 diabetes (6,7). Although the GLP-1 analog therapies may lead to loss of spatiotemporal regulation of most of GLP-1Rs, similar therapeutic benefits can be achieved with modulators of GLP-1 potency. Furthermore, because of its clean modulation without intrinsic agonist activity, the potential adverse effects are minimized. This mode of action permits control over the intensity of activation based on endogenous levels of GLP-1 and could potentially restore GLP-1 signaling function in a manner more consistent with physiological demands. As cAMP is suf-   ficient for glucose-dependent insulin release (29), future studies may investigate whether compounds that enhance GLP-1induced cAMP production represent a new class of therapeutic agents for the treatment of type 2 diabetes or other GLP-1 signaling-related disorders. To this end, examination of in vivo effects of N55 is necessary. Other studies may also investigate the viability of nutrient derivatives from edible plants in treating or preventing type 2 diabetes or other GLP-1 signalingrelated conditions.
In summary, we identified a new compound from fenugreek seeds with potential clinical application in the treatment of type 2 diabetes and other GLP-1 signaling-related disorders. N55 binds to GLP-1(7-36)-amide, enhances the potency in stimulating the cAMP pathway. This mechanism is distinct from that of classical GLP-1R-signaling agents, which directly act on the target receptor orthosterically or allosterically. N55 thus represents a new class of compounds that enhance GLP-1R signaling. Our results also suggest that GLP-1 might be a novel target for drug discovery in type 2 diabetes and other conditions. The postulated mechanism of N55 action also points to the possibility of modulating GPCR activity through modification of the potency of its cognate ligands. The approach described here is applicable to many other plant compound discovery efforts related to other GPCRs and their respective ligands.