Abstract

Introduction:

Diabetes mellitus (DM) is a multifaceted disease etiologically characterised by dysregulation in glucose homeostasis. The World Health Organization (WHO) global report indicates that over 90% of DM cases are classified as Type 2 DM (T2DM), which is clinically characterized by chronic hyperglycemia. This systemic condition arises predominantly due to the interplay of two key components: (a) compromised insulin production by the pancreatic β-cells, and (b) the failure of insulin-sensitive tissues to react to insulin. Notably, it is well established that glucagon-like peptide-1 (GLP-1), an incretin hormone of the glucagon superfamily, contributes to glucose-dependent pancreatic β-cell insulin secretion. The insulinotropic impacts of secreted GLP-1 are facilitated by its interaction with GLP-1 receptor (GLP-1R), a class B G-protein-coupled receptor (GPCR). However, GLP-1 is proteolytically cleaved by dipeptidyl peptidase 4 (DPP-4), resulting in a plasma half-life of ∼2 minutes, which limits its therapeutic efficacy in patients with T2DM. Therefore, the exogenous administration of DPP-4-resistant GLP-1R agonists (GLP-1RAs) has proven to be a successful therapeutic strategy for managing T2DM. Notably, the currently marketed GLP-1RAs, such as Semaglutide, Liraglutide, and Lixisenatide, are long-chain GLP-1 mimetic peptides, ranging in length from 33 to 39 amino acids.

Methods:

In this context, the computational design and in silico evaluation of a DPP-4- resistant, potent designer helical peptide agonist (SR18; ≤18 aa) of the GLP-1R, comprised of both coded and non-coded amino acids, are described in the current study. The basic pharmacological activity of the designer peptide, SR18 was evaluated through circular dichroism, dynamic light scattering, proteolysis, cytotoxicity and hemolytic experiments.

Results and discussions:

SR18 preserves several amino acids necessary for effective interactions with the GLP-1R, similar to those found in GLP-1, Liraglutide, and Semaglutide. Interestingly, the binding of SR18 also mimics the binding of small-molecule agonists of GLP-1R. Preliminary experimental studies confirm that synthetically prepared SR18 maintains an ordered, α-helical conformation under various solvent conditions and possesses the basic pharmaceutical properties desired of a potent lead peptide. Furthermore, compared with GLP-1 and Semaglutide SR18 exhibits stable interactions with GLP-1R over 1 μs of molecular dynamics (MD) simulations, with appreciable binding affinity and energy, supporting its viability as a potential alternative to the current long-chain GLP-1R peptide agonists.

1 Introduction

Type 2 diabetes mellitus (T2DM) is a chronic hyperglycemia-induced metabolic disease (Singh et al., 2025a), which accounts for ∼90% of patients clinically diagnosed with diabetes. There has been a notable increase in the number of patients with T2DM due to the global rise in obesity, driven by sedentary lifestyles and regular consumption of energy-dense diets (Gupta et al., 2024; Hussain, 2025). Chronic hyperglycemia contributes to insulin resistance, a metabolic condition in which the body initially fails to respond to the natural endocrine hormone insulin, secreted by the pancreatic β-cells. Eventually, in response to excessive glucose stress, the β-cells fail to produce sufficient insulin, leading to β-cell dysfunction and subsequent onset of a severe form of T2DM (Accili et al., 2025). Therefore, several classes of drugs, such as alpha-glucosidase inhibitors (AGI), amylinomimetics, biguanides, bile acid sequestrants, dopaminergic antagonists, DPP4 inhibitors (DPP4i), insulin and insulin secretagogues (IS), sulfonylureas, meglitinides, sodium-glucose cotransporter-2 inhibitor (SGLT2i), and thiazolidinediones (TZDs), including the glucagon-like peptide-1 receptor agonists (GLP-1RAs) have been made available for the management of T2DM (Tahrani et al., 2016; Drucker, 2025). However, in the process of controlling hyperglycemic conditions, antidiabetic drugs like insulin, insulin secretagogues, sulfonylureas, and meglitinides trigger hypoglycemia. Additionally, severe side effects from a few antidiabetic drugs, like TZDs, which cause congestive heart failure (Lago et al., 2007; Shi et al., 2023), and SGLT2i, which cause genitourinary infections, are also documented (Pishdad et al., 2024). Notably, among all classes of antidiabetic drugs, GLP-1RAs, a form of incretin (INtestinal secretion of INsulin) hormone-based therapy, offer additional advantages, such as inducing weight loss and preventing hypoglycemia symptoms. On the other hand, GLP-1RAs can also cause gastrointestinal side effects and injection site reactions, while carrying a higher price tag than other therapeutic options like metformin. Nevertheless, the clinical benefits of GLP-1RAs outweigh these drawbacks.

Glucagon-like Peptide-1 (GLP-1, 30–31 aa) is a natural gastrointestinal incretin hormone of the “glucagon peptide family”, secreted from the enteroendocrine L-cells in the distal gut to stimulate insulin secretion in response to postprandial hyperglycemia (Figure 1). Briefly, the preproglucagon gene undergoes tissue-specific post-translational modification with the help of specific propeptidase convertase (PC1/3) to produce two variants of GLP-1, the amidated GLP-1 [7–36], and the glycine-extended GLP-1 [7–37] in the ileum and hypothalamus (Lafferty et al., 2021; Zheng et al., 2024). However, the GLP-1 [7–36] amide accounts for 80% of circulating GLP-1 (Helmstadter et al., 2022), which lowers postprandial hyperglycemia through three established mechanisms: (a) deceleration of gastric emptying, (b) insulinotropic actions, and (c) suppression of glucagon function except for hypoglycemia (Hansen et al., 2024).

Schematic illustration of the role of GLP-1 in controlling postprandial hyperglycemia.

Notably, the insulinotropic actions of circulating GLP-1 are facilitated through a “two-domain” binding interaction with GLP-1R (Figure 2), a structurally well-characterized prototypical class B GPCR, which recruits the Gαs-subunit of the heterotrimeric G-protein for cAMP-mediated downstream signaling, involving protein kinase A (PKA) and exchange protein activated by cAMP (EPAC), leading to insulin exocytosis. Briefly, while the extracellular domain (ECD) of GLP-1R, which is responsible for ligand recognition and specificity, engages the C-terminus of the GLP-1, the interhelical transmembrane crevice of the GLP-1R helps in docking of the N-terminus of the GLP-1, triggering downstream signaling (Fang et al., 2020).

Cartoon representation of the “two-domain” binding interaction of GLP-1 with GLP-1R leading to insulin exocytosis.

However, the naturally secreted GLP-1 is quickly hydrolyzed and rendered inactive by dipeptidyl peptidase-4 (DPP-4), a transmembrane amino peptidase. This ubiquitous protease is expressed in several tissues, including pancreatic β-cells, and cleaves the amino-terminal dipeptide at the penultimate position with either L-proline or L-alanine (GLP-1), limiting the clinical utility of GLP-1 in T2DM patients, particularly its insulinotropic actions (Meier et al., 2006; Saini et al., 2023). The elimination half-life of the intact GLP-1, including the DPP-4-generated metabolites, such as GLP-1 [9–36] or GLP-1 [9–37]) is ∼2 min. Moreover, the DPP-4-generated metabolites of GLP-1 (Figure 1), lacking the His-Ala dipeptide, have ∼100-fold less affinity towards GLP-1R. DPP-4 also acts as a molecular bridge between obesity and T2DM. Several factors contribute to the expression and regulation of DPP-4 in the body. The extracellular domain of DPP-4, which harbors its catalytic site, often gets released from the membrane and circulates in plasma. The onset of obesity primes the enlarged visceral adipocytes and infiltrating macrophages to overexpress and release soluble DPP-4 in the bloodstream, where its levels correlate directly with BMI and insulin resistance in skeletal muscle cells and liver tissues. Elevated serum DPP-4 levels promote chronic inflammation, leading to the onset of T2DM (Ghorpade et al., 2018; Guo et al., 2024). Notably, some studies suggest that patients with T2DM have elevated circulating DPP-4 and lower plasma GLP-1 levels, both of which correlate positively with hyperglycemia and obesity (Lugari et al., 2004; Ryskjaer et al., 2006).

Nevertheless, the regenerative effects of GLP-1 on β-cells offer immense potential for T2DM patients, who may have lost nearly half of their β-cell activity by the time they were diagnosed (Shao et al., 2025). Therefore, GLP-1RAs that mimic the function of native GLP-1 but with improved pharmacokinetics are highly desirable as drug candidates for the management of T2DM (Prajapati et al., 2024). Thus, GLP-1 analogs are prepared considering two key points in mind: (i) Native GLP-1 is cleaved by DPP-4 at Ala8, and (ii) Arg or Lys residues are susceptible to serine protease cleavage, more specifically Lys34. Therefore, modification strategies, such as incorporation of noncoded amino acids (NCAAs), N-methylation, substitution with β-amino acids, backbone modification, and fatty acid acylation, have been employed to develop various GLP-1RAs to date (Wang et al., 2022). For instance, sequence remodeling and extension led to the discovery of exenatide (Byetta), a 39-amino acid synthetic peptide with ∼50% sequence identity to GLP-1, derived from exendin-4, a peptide found in the venom of the Gila monster Heloderma suspectum (Furman, 2012; Van Baelen et al., 2022). The substitution of Ala8 in GLP-1 by Gly8 in exendin-4 contributes to enhanced DPP-4 resistance (Movahednasab et al., 2025). Notably, various exendin-4-based GLP-1RAs (Exenatide: 39 aa and Lixisenatide: 44 aa), and GLP-1-based GLP-1RAs (Dulaglutide: 31 aa, Liraglutide: 31 aa, Semaglutide: 31 aa, Tirzepatide: 39 aa, and Albiglutide: 60 aa) have been developed and approved by the FDA between 2005 and 2022 for the management of T2DM (Liu, 2024). In addition, other GLP-1RAs, such as Beinaglutide, Supaglutide, Emaglutide, and Efpeglenatide, are currently in clinical trials.

Interestingly, until now, primarily long-chain peptides with 33–39 amino acids have been developed as GLP-1RAs, which also carry a cost burden. Notably, non-peptide small-molecule agonists of the GLP-1R are also being developed for oral bioavailability (Knudsen et al., 2007; Kawai et al., 2020; Griffith et al., 2022). Therefore, there should be a reasonable effort to design alternative short-chain peptide-based GLP-1RAs with low toxicity and efficacy comparable to marketed long-chain peptides. In fact, designer peptides composed of nine to eleven amino acids have been developed earlier as potent GLP-1RAs that exhibit affinity and potency similar to the native GLP-1 (Mapelli et al., 2009; Jazayeri et al., 2017). Therefore, short peptides containing crucial amino acids required for interaction with GLP-1R can be designed to act as small-molecule GLP-1RAs for insulin exocytosis under hyperglycemic conditions. In this context, using a previously reported α-helical peptide as a template sequence (Behera et al., 2022), this study discusses the rational computational design and in silico evaluation of an α-helical peptide (SR18: 18 aa) as a potent agonist of GLP-1R, with native GLP-1 and Semaglutide serving as references. The analysis of molecular dynamics (MD) simulation data over 1 μs each for the GLP-1–GLP-1R and the SR18–GLP-1R systems, estimation of free energy of binding, including preliminary experimental studies on the synthetically prepared designer peptide, suggests that the designer peptide SR18 has strong signature traits of an agonist that can potentially trigger GLP-1R-mediated insulin exocytosis under experimental conditions in tissues.

2 Materials and methods2.1 General computational methods

Protein Data Bank (PDB) coordinates of the GLP-1−GLP-1R (5VAI) (Zhang et al., 2017), Semaglutide (7KI0), Taspoglutide (7KI1) (Zhang et al., 2021), and liraglutide (4APD) were retrieved from the RCSB PDB server (www.rcsb.org). Multiple sequence alignments of the peptides were performed through Clustal Omega (Sievers and Higgins, 2018; Madeira et al., 2024). MODELLER (Webb and Sali, 2016), Discovery Studio (Biovia), and PyMOL (Schrӧdinger, Inc.) software were utilized for modeling, initial processing, visualization, analysis, and presentation of GLP-1R and model peptide structures. The data were plotted in GraphPad Prism (www.graphpad.com).

2.2 Molecular docking studies

The HDOCK web server (Yan et al., 2017b; Yan et al., 2020) was used to perform peptide-protein docking to predict the putative binding mode between the designer peptides and loop remodeled GLP-1R. HDOCK is based on a hybrid algorithm that combines template-based modeling and ab initio free docking. The PDB files of modeled GLP-1R and peptides were uploaded to the HDOCK server, and global docking was performed using the default parameters, which use a fast Fourier transform (FFT)-based rigid-body search method and then evaluate the interface energetics using an intrinsic scoring function (Huang and Zou, 2008; 2014; Yan et al., 2017a). For each docking run, the server generated the top 100 predicted complex structures, along with their HDOCK energy scores and confidence scores. Docking energy scores are calculated using iterative scoring functions ITScorePP or ITScorePR. This energy score indicates the expected binding affinity between the ligand and the receptor; more negative scores indicate higher binding affinity. The confidence score provides an empirical measure of binding likelihood between docked molecules. Confidence score >0.7 suggests strong binding, whereas a value <0.5 indicates poor binding (Li et al., 2022). The best conformation complex (with the highest HDOCK energy score) of the GLP-1R and the peptides was further screened for MD simulation. Additionally, the binding energies and the ∼Kd values of the top docking poses obtained from HDOCK were also estimated using PRODIGY (Xue et al., 2016).

2.3 Peptide design and synthesis

The sequence of the GLP-1 peptide and a few reported GLP-1RAs, like Exendin-4, Semaglutide, Liraglutide, and Taspoglutide, were first aligned to obtain the mutating residue. Further, one of the earlier reported model helical peptides (Hed3: NH3+-YGKAAAAK-Igl-AAAKAAAAK-CO2−) was utilized to generate the template GLP-1 mimetic peptide (Hed6: NH3+-H-Igl-E-Aib-AAAKAAAAKAAAKA-CONH2), which was subsequently subjected to iterative computational sequence optimization. The 3D structure of the helical peptides was generated using the Discovery Studio Visualizer tool. Subsequently, one of the rationally designed peptides (SR18) was custom-synthesized using the standard Fmoc (Fluorenylmethoxycarbonyl protecting group) solid-phase peptide synthesis method by availing the commercial services of GenScript Biotech. Post cleavage from the resin, the peptide was purified (≥95%) to homogeneity. The analytical high-performance liquid chromatography (HPLC) profile of the peptide was recorded by using a C18 (4.6 × 250 mm2) column at 220 nm under an acetonitrile-water gradient in the presence of 0.05%–0.065% trifluoroacetic acid (TFA). Electrospray ionization mass spectrometry (ESI-MS) confirmed the integrity of the peptide.

2.4 Circular dichroism (CD) studies

The far-UV circular dichroism (CD) spectra of the SR18 were recorded in a quartz (Helma) cell with a path length of 0.1 cm at 25 °C under different solvent conditions, such as 1X PBS (phosphate buffered saline), 30 mM SDS (Sodium dodecyl sulphate), 20% TFE (2,2,2-trifluroethanol), and 20% HFIP (1,1,1,3,3,3-hexafluro-2-propanol), by using the Chirascan CD spectrometer system (Applied Photophysics). The CD spectra were recorded with a time constant of 1 s and a step size of 1 nm, averaging three scans. The peptide stock solution was diluted in the respective solvents, and the peptide concentration was maintained at 100 μM by monitoring the absorbance at 280 nm. After background subtraction of corresponding solvents, the observed ellipticity in millidegrees was converted to mean residual ellipticity [θMRE, degcm2·dmol-1] using the following equation (millidegrees × mean residual weight)/(path length in millimeters × concentration in mg·mL-1) (Greenfield, 2006). The ⍺-helix fraction (fa) of SR18 at θ222 was calculated as the ratio of [θ222]experimental to [θ222]theoretical, as described here (Hamley et al., 2025). All data points were plotted in GraphPad Prism, and the spectra were smoothed up to 2–3 units wherever required.

2.5 Dynamic Light Scattering (DLS) studies

SR18 dissolved in filtered 1X PBS (pH ∼ 7.4) and 20% TFE to a final concentration of 100 μM was subjected to DLS measurement to gauge its hydrodynamic size and particle size distribution, using bovine serum albumin as the reference. Prior to the DLS measurement, the sample was filtered through a 0.22 μm filter to eliminate the dust particles that could interfere with the measurement. The analysis was performed at 25 °C using a Zetasizer Nano ZS (Malvern Instruments) equipped with a 633 nm He-Ne laser, with scattered light detected at 90°. Each measurement consisted of three independent runs, each averaging five acquisitions.

2.6 Cytotoxicity studies

The cytotoxicity profile of SR18 was examined using the 3-(4,5-dimethylthiazol-2-yl) 2,5-diphenyltetrazolium bromide (MTT) assay utilizing HEK-293 and HeLa cell lines obtained from the School of Biological Sciences, NISER, Bhubaneswar. Approximately 4 × 104 cells per well were inoculated into 96-well plates containing Dulbecco’s modified Eagle’s medium (DMEM) enriched with 10% fetal bovine serum and penicillin-streptomycin (10 mL/L), and incubated for 24 h at 37 °C with 5% CO2. Thereafter, the cells were treated with serially diluted SR18 in culture media and incubated for an additional 18 h at 37 °C in 5% CO2. Subsequently, 10 µL of MTT solution (5 mg/mL) was introduced to each well, and the plates were incubated for a further 4 h. After 4 h, the suspensions were removed, and 100 µL of dimethyl sulfoxide (DMSO) was added to dissolve the formazan crystals. The plates were further incubated at 37 °C with 5% CO2 for 30 min to achieve complete solubilization. The absorbance was measured at 570 nm using a SpectraMax iD3 microplate reader (Molecular Devices). Cells treated with PBS served as the negative control, whereas cells exposed to 10% DMSO served as the positive control. After subtracting blank absorbance (medium only), cell viability (%) was calculated as the ratio of OD570 of treated cells to OD570 of untreated cells. To ensure reproducibility, all experiments were conducted in triplicate and repeated three times independently (n = 3).

2.7 Hemolytic Assay

Approximately 2 mL of human blood was collected from an anonymous healthy volunteer at the Sanjeevan Health Center, Indian Institute of Technology Bhubaneswar, following written informed consent. Blood was collected in a K2-EDTA-coated tube to prevent coagulation. Erythrocytes were separated by centrifugation at 2,500 rpm for 10 min at room temperature. The collected erythrocytes were then washed four to five times with sterile 1X PBS, with each wash including 10 min of centrifugation, until the supernatant was clear. The purified erythrocytes were then diluted in 1X PBS to a final ratio of 1:50 (Erythrocytes: 1X PBS). Aliquots of the diluted erythrocyte suspension were introduced into microcentrifuge tubes containing 10 μL of SR18 with concentrations ranging from 0.39 to 100 μM, resulting in a final reaction volume of 100 μL. The mixtures were incubated at 37 °C for 60 min, then centrifuged at 2,500 rpm for 10 min at room temperature. After that, ∼50 μL of supernatant was carefully transferred to microtiter plate wells without disturbing the cell pellet, and the cell lysis of the erythrocytes was quantified by measuring OD at 414 nm using SpectraMax iD3 microplate reader (Molecular Devices). Cells treated with 10 μL of 10% Triton X-100 served as the positive control, while cells treated with 10 μL of sterile PBS served as the negative control. All experiments were carried out in triplicate on two distinct experimental sets (n = 2). The proportion of hemolysis was estimated using the formula (Abspeptide-AbsPBS)/(Abstriton-AbsPBS) × 100.

2.8 Proteolysis studies

SR18 was incubated with trypsin at an enzyme-to-peptide ratio of 1:20 (trypsin: peptide) in 50 mM Tris−HCl (pH ∼ 8) at 37 °C with moderate agitation for ∼4 h to evaluate its susceptibility to proteolytic degradation. Subsequently, the reaction mixture was heated to 100 °C for 30 min to inactivate trypsin, then centrifuged at 13,000 g for 30 min to precipitate the enzyme. The resultant supernatant was collected and analyzed using an Exactive Plus Orbitrap liquid chromatography-high-resolution mass spectrometry (LC-HRMS) instrument from Thermo Scientific.

2.9 Molecular dynamics studies

The GLP-1R, respectively complexed to GLP-1 and SR18, was embedded into a 75% POPC (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine) and 25% cholesterol (CHL) bilayer using CHARMM-GUI (Jo et al., 2008) before MD simulation. The GLP-1−GLP-1R and SR18−GLP-1R complexes were subjected to MD simulations of 1 μs at 300 K in TIP3 water using CHARMM36 m forcefield (Huang et al., 2017) parameters in the GROMACS engine (Hess et al., 2008), as described in our previous studies (Gupta et al., 2025). Before the production MD run, both systems were energy-minimized to a tolerance of 1,000 kJ mol-1nm-1 using the steepest descent method, first in a vacuum and then in the presence of TIP3 water, with a solvent density set to the value corresponding to 1 atm at 300 K. Furthermore, the net charge of the system was neutralized by randomly placing the required number of counterions in the box to maintain a NaCl concentration of 0.15 M. The final GLP-1−GLP-1R complex contained 72,067 atoms, including 16,320 TIP3 water, 44 sodium, 99 chloride, 102 POPC, and 33 CHL molecules, whereas the SR18−GLP-1R complex contained 79,689 atoms, including 18,915 TIP3P water, 51 sodium, 54 chloride, 102 POPC, and 33 CHL molecules. Both systems were equilibrated twice, first under constant number, volume, and temperature (NVT: 5 ns) conditions, and then under constant number, pressure, and temperature (NPT: 10 ns) conditions before the start of the production MD run. With a coupling time constant of 1 ps, the protein and solvent were coupled independently to a V-rescale bath at 300 K. H-bonds were constrained with LINCS with order 4. The non-bonded pair list cutoff was 1.2 nm with a grid function. Numerical integrations were performed in steps of 2 fs, and the coordinates were updated every 10 ps. Conformational clustering of both the trajectories (SR18−GLP-1R and GLP-1−GLP-1R) was performed every 50 ps with a root mean square deviation (RMSD) cutoff of ≤2.5 Å by recruiting the GROMOS fitting method, as defined in GROMACS. The MD trajectories were thoroughly analyzed by recruiting the utility modules available in GROMACS. A similar process was followed for the MD studies of GLP-1, Semaglutide, and SR18 in the presence of TIP3 water. The MD trajectories of the peptides were subjected to conformational clustering at 50 ps intervals using an RMSD cutoff of ≤1.5 Å.

2.10 Estimation of the free energy of binding

The following equation, ΔGbinding = Gcomplex −(Gprotein + Gligand), implemented in the gmx_mmpbsa program (Valdes-Tresanco et al., 2021), was used to calculate the binding free energy of the complexes using the Molecular Mechanics Poisson-Boltzmann surface area (MM-PBSA) method. The following equation estimated the free energy contribution: G = EMM + G(solv) – TS(solute), where EMM (molecular mechanics energy) represents the summation of van der Waals and electrostatic, G(solv) represents the solvation energy contributed by both polar and non-polar solvation free energy, and TS(solute) represents the temperature and entropy of the solute. Finally, 300 structures, including 200 conformers randomly selected from the first major cluster and 100 from the second major cluster, were used from the respective MD trajectories to calculate the total binding free energy and decomposition energy of the complexes.

3 Results3.1 Molecular dynamics (MD) simulation of GLP-1 complexed to loop remodeled GLP-1R

The initial structure of GLP-1R was retrieved from the reported active structural complex of GLP-1R with the native GLP-1 (PDB: 5VAI). The unresolved loop residues (S129RRGEK134) connecting the N-terminus to the transmembrane helix 1 (TM1) of GLP-1R were appropriately remodeled using MODELLER. The native GLP-1 was then docked to the loop-remodeled GLP-1R using the HDOCK web server, which generated a GLP-1R complex with GLP-1 with a docking energy score of −653.61 and a confidence score of 1. The structural superposition of the modeled complex with 5VAI indicated a backbone RMSD of 1.079 Å (Figure 3), suggesting that the model complex is suitable for further studies.

Docking of native GLP-1 (green) to the loop-remodeled GLP-1R with structural superposition against the cryo-EM reported structure of GLP-1 activated GLP-1R (PDB: 5VAI). The amino acid side chains of the remodeled loop region are shown in pink sticks.

The model GLP-1–GLP-1R complex was embedded in a POPC-CHL (75:25) bilayer using CHARMM-GUI and subsequently subjected to MD simulations for 1 μs at 300 K as a positive control. The MD data presented in Figure 4 suggest that the GLP-1–GLP-1R complex remains structurally stable throughout the MD simulations, with an average of 15 intermolecular hydrogen bonds between GLP-1 and GLP-1R. Further analysis of the MD trajectory reveals that the central conformer of the most populated cluster of the GLP-1−GLP-1R complex evolved over 1 μs of MD simulations, shares an RMSD of 3.057 Å (Figure 4) with the starting model complex. Notably, the remodeled section of the loop, which was originally not resolved in the cryo-EM studies, adopted an altogether different conformation in the most populated conformer of the complex.

(a) The energy-minimized model complex of GLP-1–GLP-1R, embedded in the model POPC + CHL bilayer. (b) Structural superposition (RMSD: 3.057 Å) of the GLP-1–GLP-1R complex with the central conformer of the most populated cluster of the complex, as evolved over 1 μs of MD simulations. (c) Major conformational clusters of the GLP-1–GLP-1R complex evolved over 1 μs of MD simulations. (d) The overall structural stability of the GLP-1–GLP-1R complex over the duration of the MD simulations.

Further analysis of the trajectory suggests that the several important amino acids, such as H7, E9, T11, F12, D15, and F28 of GLP-1 actively interact with the W39, Y145, Y152, R190, K197, L201, N300, R310, E364, L384, and L388 of GLP-1R through several hydrogen, hydrophobic, and electrostatic interactions (Figure 5; Supplementary Figure S1). Interestingly, H7, E9, and D15 of GLP-1 destabilize TM6 and ECL3, thereby facilitating the GLP-1R in attaining its active conformation. Further, the H7 of GLP-1 interacts with TM3 and TM5 through hydrogen and hydrophobic interactions with R310, I313, and E364 of GLP-1R. Notably, earlier alanine substitution studies have revealed that Y152, R190, K197, L201, R299, N300, W306, R310, I313, E364, L384, L388 of GLP-1R are the most important residues for GLP-1 binding affinity and potency. Interestingly, except for Y148, D198, M233, Q234, Y235, W284, D372, E373, R380, K383, and E387, all other residues were found to maintain sustained intermolecular interactions with GLP-1 over the duration of the MD simulations (Dods and Donnelly, 2015; Liang et al., 2018; Zhang et al., 2020; Deganutti et al., 2022). Therefore, it is reasonable to assume that GLP-1 analogs that could establish stable interaction with these residues of GLP-1R can produce therapeutic effects similar to the current GLP-1R targeting peptides for the management of T2DM.

Monitoring the sustainability of intermolecular (hydrogen, electrostatic, and hydrophobic) interactions observed between GLP-1 and GLP-1R throughout the MD simulation. The GLP-1R residues are shown in superscript. The grey line highlights the cutoff distance for the specified interaction between the amino acids.

3.2 Rational design of an α-helical peptide agonist for targeting GLP-1R

Sequence comparison of GLP-1 (30 aa) with the other GLP-1 mimetics, such as Taspoglutide (30 aa), Liraglutide (31 aa), and Semaglutide (31 aa), indicates that strategic incorporation of NCAAs, such as Aib, and chemical modification of specific Lys side chains have been successful in improving the pharmacokinetics of the GLP-1 mimetics without affecting their interactions with GLP-1R compared to the native GLP-1. Further, analysis of the reported cryo-EM data of GLP-1–GLP-1R complex indicates that certain residues on GLP-1 do not make significant contact with the ECD of the GLP-1R (Zhang et al., 2017). Therefore, for rational design of peptide mimetics, the entire sequence of GLP-1 may be divided into the N-terminal (14 aa), connector (5 aa), and C-terminal (11 aa) regions, which are collectively important for binding and signaling of GLP-1R. Since short peptides ranging from nine to eleven amino acids have been discussed in the literature earlier as potent GLP-1RAs, we decided to design an intermediate-sized potent peptide (≤20 aa) agonist of GLP-1R by conserving several critical residues of GLP-1 important for GLP-1R binding and signaling. Notably, in our earlier studies, we demonstrated the impact of a few NCAAs on the conformational stability of model helical peptides composed of ≤18 amino acids (Behera et al., 2022), which we subsequently subjected to sequence optimization, producing a potent antimicrobial peptide (Behera et al., 2024a) and a potential antiviral peptide (Behera et al., 2024b) binder targeting the receptor binding domain of SARS-CoV-2.

Therefore, in the current study, we utilized one of the model helical peptides (Hed3: NH3+-YGKAAAAK-Igl-AAAKAAAAK-CO2−) to produce the template GLP-1 mimetic peptide (Hed6: NH3+-H-Igl-E-Aib-AAAKAAAAKAAAKA-CONH2) and further subjected it to iterative computational sequence optimization to produce a potent helical peptide agonist (SR18) of GLP-1R (Figure 6). The Lys residue at the 17th position was introduced at the C-terminus with an objective to modify its sidechain with a linker in the future (Supplementary Figure S2), so that the peptide can anchor to serum albumin similar to Semaglutide (Knudsen and Lau, 2019). Additionally, the C-terminus was amidated to protect the peptide from carboxypeptidase degradation.

Schematic illustration of the sequence optimization process utilized in the rational design and selection of SR18. Residues highlighted in green represent the critical conserved residues. The red-colored amino acids represent NCAAs (Igl, X = Aib). The lysine residue, which is chemically modified in GLP-1 analogs, is highlighted in orange.

The first four residues of the N-terminus were also fixed for the following reasons. His1 and Glu3 were fixed as critical residues for GLP-1R (Adelhorst et al., 1994; Xiao et al., 2001) Further, whereas Igl (Indanyl glycine), a glycine derivative, was introduced at the second position to protect the dipeptide cleavage by DPP-4 and to improve metabolic stability, including imparting necessary conformational constraints to maintain an appreciable binding affinity at the active site of GLP-1R, similar to the small-molecule GLP-1RAs. Furthermore, the Aib at the fourth position was fixed to help the peptide induce a helical turn at the N-terminus (Karle and Balaram, 1990). Notably, Aib has been incorporated into GLP-1RAs, such as Taspoglutide, Semaglutide, and Tirzepatide, to improve their therapeutic index. The remaining amino acid positions were iterated to produce ∼200 unique peptides, each composed of ≤18 amino acids, out of which 41 peptides were initially shortlisted and ranked (Supplementary Table S1) through screening against loop-remodeled GLP-1R bound to GLP-1 based on their binding energy score and orientation using the HDOCK webserver.

The binding energy scores of the peptides ranged between −352.98 (top rank, confidence score: 0.98) and −202.07 (bottom rank, confidence score: 0.73), compared to GLP-1 (−653.61, confidence score: 1.00), with higher negative values indicating relatively stronger binding affinity. Furthermore, a confidence score >0.7 indicates a higher likelihood of binding to the receptor. Subsequently, 22 peptides (∼50% of the shortlisted peptides) with a confidence score ≥0.92 were superimposed in the context of GLP-1 bound to GLP-1R (Figure 7). Notably, the docking orientation of the top-ranked peptide was observed to be different compared to that of the GLP-1. Interestingly, exendin-4 (9–39), an antagonist of GLP-1R (Avexitide, Phase III), has been shown earlier to bind to the ECD of GLP-1R with an orientation different than that of GLP-1. Notably, activation of GLP-1R requires specific coordinated interactions with potential ligands, as observed in GLP-1 and Semaglutide. Therefore, all high-affinity binding ligands may not necessarily behave as GLP-1RAs and may potentially exhibit alternative functions, such as biased agonists, inverse agonists, and antagonists. In fact, it is reported (Underwood et al., 2010) that exendin-4 (9–39) binds strongly (EC50 ∼5 pM, IC50 ∼0.76 nM) to wild-type GLP-1R than the native GLP-1 (EC50 ∼11 pM, IC50 ∼1 nM). Moreover, the top-ranked peptide did not contain the amino acids in its sequence considered important for activating the GLP-1R and was thus reserved for future evaluations.

(a) Structural superposition of the designer peptides in the context of GLP-1 bound to the modeled GLP-1R. (b) Comparison of the binding orientation of the top-ranked peptide (red, binding score: −352.98) with the SR18 (blue, binding score: −272.14) in the context of GLP-1 (green, binding score: −653.61). The SR18 demonstrates a backbone RMSD of 0.948 Å with GLP-1, whereas the top-ranked peptide demonstrates a backbone RMSD of 1.094 Å with GLP-1 and an RMSD of 0.721 Å with SR18. (c) Illustration of various intermolecular interactions observed in the docking pose between SR18 and GLP-1R. The green line indicates hydrogen bonding, the pink line indicates hydrophobic interactions, and the orange line indicates electrostatic interactions.

However, another peptide, codenamed SR18 (H-Igl-E-Aib-TFTSDLSDHYESKA-CONH2), which contained several amino acids similar to those found on the GLP-1 and demonstrated a docking orientation similar to the GLP-1 with a binding energy score of −272.14, was rationally shortlisted for further studies as a first-generation GLP-1RA over the top-ranked peptide (H-Igl-E-Aib-AWMTSIISTMWAKH-CONH2). Interestingly, SR18 also docked to the most populated conformer of the loop-remodeled GLP-1R bound to GLP-1, evolved over 1 μs of the MD simulations in a POPC-CHL bilayer. Notably, SR18 occupied the same orthosteric binding pocket on both conformers of GLP-1R, a binding site typically occupied by GLP-1 and other similar peptide mimetics, as evidenced in reported structural studies (Zhang et al., 2017; Zhang et al., 2020). A comparative docking energy (confidence score) and binding energy for SR18 in reference to GLP-1, Semaglutide, and Avexitide are provided in Table 1.

PeptideReceptorHDOCK energy (confidence) scorePRODIGY energy (Kd)GLP-1GLP-1R (0 ns)−653.61 (1.00)−12.9 kcal/mol (0.49 nM)GLP-1R (central conformer)−328.81 (0.97)−12.2 kcal/mol (1.2 nM)SemaglutideGLP-1R (0 ns)−478.56 (0.99)−12.8 kcal/mol (0.44 nM)a