Protease-Resistant Azapeptide GLP-1 Analogue Improves Metabolic Control in Diet-Induced Obesity

Abstract

Peptide therapeutics are limited by rapid proteolysis and short half-lives. Azapeptides, created by replacing one or more α-carbon(s) on the peptide backbone with nitrogen atom(s), offer a strategy to improve peptide stability while preserving functional efficiency, yet their clinical potential has remained underexplored. Herein, we report the design, synthesis, in vitro and in vivo evaluations of azapeptide-based glucagon-like peptide-1 receptor agonists (GLP-1RAs). Using a solid-phase synthesis platform, we generated GLP-1 analogues with aza-substitutions at protease-sensitive residues. The lead analogue, AzaA⁸/R³⁴-GLP-1(AzaA8), resisted dipeptidyl peptidase-4 degradation (>24 h), maintained picomolar potency at the GLP-1 receptor (GLP-1R) signaling, and exhibited an extended plasma half-life in mice relative to unmodified controls. In lean mice, AzaA8 improved oral glucose tolerance, and in high-fat diet-induced obese mice, chronic administration reduced body weight, decreased leptin and insulin levels, and enhanced glucose handling without detectable inflammatory adverse effects. These findings demonstrate that a targeted aza-substitution yields a protease-stable, biologically active GLP-1RA with metabolic benefits, establishing azapeptides as a promising scaffold for next-generation incretin-based therapies in diabetes and obesity.

Introduction

Obesity and type 2 diabetes are escalating global health crises that drive cardiovascular disease, fatty liver, and other metabolic complications¹. Current therapies often show limited long-term efficacy, side effects, or require frequent administration, underscoring the need for safer, longer-acting interventions^2–4. One successful class of treatments is the incretin-based peptide therapeutics, exemplified by glucagon-like peptide-1 receptor agonists (GLP-1RAs), which enhance insulin secretion and suppress appetite⁵.

GLP-1RAs such as semaglutide and liraglutide (Fig. 1a, 1d) have transformed the management of diabetes and obesity by improving glycemic control and inducing weight loss. However, native GLP-1 is rapidly inactivated by dipeptidyl peptidase-4 (DPP-4; Fig. 1b), and even clinically used analogues require extensive chemical modifications to modestly extend half-life⁶. For example, Ozempic^® (semaglutide) retains the GLP-1(7–37) scaffold but incorporates an Ala8→α-aminoisobutyric acid (Aib) substitution to block DPP-4 cleavage, a Lys34→Arg substitution, and an 18-carbon fatty diacid attached at Lys26 to confer albumin binding and slow clearance. While these modifications are effective, they add complexity and highlight the need for alternative strategies to engineer peptides with intrinsic stability.

Azapeptides, in which α-carbon(s) is replaced by nitrogen(s) in the peptide backbone, offer one such strategy. This conservative substitution preserves the overall backbone geometry while rendering the modified bond resistant to protease cleavage^7,8. Despite their potential to increase stability and half-life, azapeptides have seen limited clinical translation [goserelin (Zoladex^®) is the sole FDA-approved example], in part due to historical synthetic challenges⁹. We recently developed a platform method that enables efficient incorporation of pre-activated aza-amino acid building blocks into peptides using standard solid-phase synthesis^10–13, opening the door to explore azapeptide therapeutics more broadly.

Here, we applied this approach to a therapeutically relevant peptide hormone, designing GLP-1 analogues with aza-substitutions at known protease-sensitive positions. We identified a lead candidate, azaA8 (Ala8→aza-Ala in an Arg34-GLP-1 background), that resists DPP-4 degradation, retains low-picomolar receptor agonist potency, and produces significant metabolic benefits in both lean and diet-induced obese mice. These findings demonstrate that aza-substitution is a generalizable strategy for generating stable, efficacious incretin mimetics with translational potential for the treatment of diabetes and obesity.

Results

Azapeptide design and DPP-4 stability

To evaluate whether backbone aza-substitution(s) can stabilize a clinically relevant peptide hormone, we synthesized GLP-1(7–37) analogues bearing N-terminal aza-residues using our solid-phase platform¹³. The parent sequence was modified with a Lys34→Arg substitution (R34-GLP-1, as in semaglutide). We generated analogues with aza-substitutions at His7, Ala8, both His7 and Ala8 (with Ala8 optionally replaced by Aib in one variant), and at Glu9 (Fig. 1c–e), and benchmarked against unmodified R34-GLP-1, “unlipidated semaglutide” (Aib8), marketed GLP-1RAs (semaglutide, liraglutide, and exendin-4). All peptides were synthesized and purified using our standard solid-phase methods with aza-amino acid building blocks, and their identities and purities were confirmed by mass spectrometry (Supplementary Table 1)¹³.

In recombinant DPP-4 digestion assays, R34-GLP-1(7–37) was degraded within minutes (half-life, t_½ ≈ 0.3 h). The protective Aib8 substitution in the unlipidated semaglutide control extended its half-life to ~5.4 h (Table 1 and Supplementary Table 2), whereas, all aza-modified analogues remained intact beyond 24 h, with no measurable loss after 48 h. Thus, replacement of His7, Ala8, or adjacent residues with aza-amino acids completely abolished cleavage at the canonical DPP-4 site, conferring a level of protection surpassing current clinical modifications.

Table 1. Composition, stability, pharmacokinetics, and receptor potency of GLP-1 analogues.

Recombinant DPP-4 stability is reported as half-life at 37 °C (t½, DPP-4, h); plasma pharmacokinetic half-life (t½, pK, h) was determined after a single IV dose (1 mg kg⁻¹) in male ICR mice; GLP-1R potency is EC_₅₀ (pM) from a CRE-Luc assay in BHK cells expressing human GLP-1R. Data are mean ± SD (DPP-4 and EC_₅₀: three independent experiments; pK: n = 3 mice per group).

Peptide	Abbreviation	In vitro DPP-4 t½ (h)	In vitro receptor activity cAMP assay (EC50, pmol l−1)	In vivo plasma t½ (pK) (h)
[R34]-GLP-1(7–37)	R34-GLP-1	0.3 ± 0.1	11.3 ± 1.2	0.08 ± 0.01
[Aib8/R34]-GLP-1(7–37)	Aib8 (unlipidated semaglutide)	5.4 ± 0.6**	8.84 ± 1.2ns	0.32 ± 0.09*
[azaAla8/R34]-GLP-1(7–37)	AzaA8	>24	22.3 ± 2.4**	0.46 ± 0.12*
[azaHis7/R34]-GLP-1(7–37)	AzaH7	>24	743 ± 164*	1.67 ± 0.52*
[azaHis7/Aib8/R34]-GLP-1(7–37)	AzaH7/Aib8	>24	306 ± 44**	ND
[azaHis7/azaGlu9/R34]-GLP-1(7–37)	AzaH7/AzaE9	>24	> 10,000	ND

Abbreviations/notation: Aib, α-aminoisobutyric acid/α-methylaniline; aza, α-carbon→nitrogen substitution; R34-GLP-1, Lys34→Arg variant of GLP-1(7–37); Aib8 (unlipidated semaglutide), Aib⁸/R³⁴-GLP-1; AzaA8, azaAla⁸/R³⁴-GLP-1; AzaH7, azaHis⁷/R³⁴-GLP-1; AzaH7/Aib8, azaHis⁷/Aib⁸/R³⁴-GLP-1; AzaH7/AzaE9, azaHis⁷/azaGlu⁹/R³⁴-GLP-1. “>24 h” indicates no measurable degradation within the assay window; ND, not detected.

Statistics are unpaired, two-tailed Welch’s t-tests versus R34-GLP-1:

^ns, not significant

^*P < 0.05

^**P < 0.01.

In vivo proteolysis shifts to internal hotspots

To confirm that N-terminal aza-substitution protects the canonical DPP-4 site in a physiological context, we performed tandem mass spectrometry (MS/MS) fragmentation analysis of three analogues (Aib8, AzaA8, and AzaH7) in mice. Peptides were administered intravenously and fragments in whole blood and urine were identified by UPLC–MS/MS over time. All three aza-GLP-1 analogues showed a common pattern of cleavage primarily within the central region of the peptide. A heat map of peptide bond cleavages (Fig. 2a, b) revealed recurrent “hotspots” around bonds 12, 14, 17, 19–21, 23–24, and 28–29, corresponding to the sequence FTSDVSSYLEGQAAKEFIAWLVR. In contrast, fragmentation at the extreme N-terminus was greatly reduced. The signature DPP-4 cleavage between positions 7–8 (His-Ala in the native peptide) was notably absent from the dominant fragments. Instead, only low-intensity signals corresponding to the N-terminal dipeptide or its immediate flanking fragments were detected, while high-intensity ions were derived from internal breaks in the peptide (Fig. 2c–e and Supplementary Datasets 1–4). These data confirm that the aza- (or Aib-) substitution at the His7/Ala8 site effectively blunted cleavage at this locus in vivo, causing proteolysis to shift to downstream sites. This site-specific protection suggests that future designs might further improve stability by reinforcing the internal cleavage-prone regions (e.g., around bonds 17–21 and 23–29) identified in the spectra. In summary, MS/MS mapping demonstrates that aza-substitution at the DPP-4 hotspot provides robust protection against N-terminal degradation, with proteolytic breakdown converging on shared internal sites across the different analogues.

AzaA8 preserves GLP-1 receptor potency

We next assessed whether the aza modifications affected the ability of the peptides to activate the GLP-1 receptor. A cell-based luciferase reporter assay (CRE-Luc) in BHK cells expressing human GLP-1R was used to determine half-maximal effective concentrations (EC₅₀) for cAMP signaling (Table 1; Supplementary Table 3). R34-GLP-1 and Aib8 controls had EC₅₀ values of 11.3 ±1.2 pmol l⁻¹ and 8.8 ±1.2 pmol l⁻¹, respectively, consistent with their high potency. AzaA8 exhibited an EC₅₀ of 22.3 ±2.4 pmol l⁻¹, indicating that substituting Ala8 with aza-Ala preserved nearly the full agonist potency of the peptide. In contrast, analogues containing an aza-His7 substitution showed somewhat reduced activity: EC₅₀ values were in the mid-to-high picomolar or low nanomolar range (e.g., 0.3–0.7 nmol l⁻¹ for AzaH7/Aib8 and AzaH7, and >10 nmol l⁻¹ for the doubly substituted AzaH7/azaE9). These results identified AzaA8 as the optimal balance of stability and activity, fully resistant to DPP-4 yet retaining low picomolar GLP-1R agonism (Table 1).

Pharmacokinetics reveal extended half-life

We next evaluated how aza-substitutions (AzaA8 and AzaH7) influenced the pharmacokinetic behavior of the peptides in vivo. Male CD-1 mice received a single intravenous dose of each peptide (1 mg kg⁻¹), and plasma concentrations were monitored over 8 h (Table 1; Supplementary Table 4). Unmodified GLP-1(7–37) was rapidly cleared from circulation, with a plasma half-life of approximately 0.08 hours (4.8 minutes), consistent with the peptide’s known instability. Introduction of the Aib substitution at position 8, which is used in semaglutide, extended the plasma half-life to about 0.32 hours (19.2 minutes). Our aza-engineered peptide AzaA8 displayed a further increased half-life to approximately 0.46 hours (27.6 minutes), while AzaH7 persisted for markedly longer, with a half-life of around 1.67 hours (100.2 minutes).

The extended persistence of the aza-analogues in circulation confirms that these substitutions slow peptide clearance in vivo. Notably, AzaA8 displayed a longer half-life yet a higher plasma clearance rate (~43.9 ml min⁻¹·kg⁻¹) than the Aib8 control (~22.6 ml min⁻¹·kg⁻¹). This apparent paradox suggests that although AzaA8 resists enzymatic degradation, it may be removed more rapidly through alternative pathways, such as renal filtration, or may have a reduced volume of distribution. By contrast, AzaH7 showed a markedly lower clearance (~7.0 ml min⁻¹·kg⁻¹), consistent with its prolonged circulation time. Overall, incorporation of aza-residues at the peptide N-terminus substantially extended systemic levels relative to native GLP-1, underscoring the stabilizing effect of these modifications.

Acute glucose lowering in lean mice

To determine whether the enhanced stability of AzaA8 translated into functional efficacy, we tested it in an oral glucose tolerance test (oGTT). Male 21-week-old C57BL/6N mice, maintained on standard chow, were fasted for 6 h with free access to water before the experiment. Groups of weight-matched animals (n = 8 per group) received a single subcutaneous injection at the back of the neck of AzaA8 (0.125 or 0.4 mg kg⁻¹), the unlipidated semaglutide analogue Aib8 (0.125 or 0.4 mg kg⁻¹), exendin-4 (0.125 mg kg⁻¹, positive control), or saline vehicle (Fig. 3a). All test articles and controls were blinded to the investigators, and mice were monitored for body weight and clinical condition (Supplementary Table 5). Two hours after injection, mice underwent an oral glucose tolerance test: each animal was given a glucose bolus (2 g kg⁻¹, delivered as a 100 mg ml⁻¹ solution in sterile water) by oral gavage, and blood glucose was measured from tail vein samples at 0, 15, 30, 60, 90, and 120 min.

AzaA8 significantly reduced glycemic excursion in a dose-dependent manner (Fig. 3b, c, Supplementary Fig. 1, and Supplementary Table 6). At the higher dose of 0.4 mg kg⁻¹, AzaA8 nearly matched the glucose-lowering efficacy of Aib8 and approached that of exendin-4. Even at 0.125 mg kg⁻¹, AzaA8 blunted the glucose rise relative to saline-treated controls. These findings demonstrate that the stability conferred by aza-substitution translates into preserved biological efficacy in vivo, enabling AzaA8 to provide acute glycemic control comparable to clinically validated GLP-1 receptor agonists like Exendin-4.

Sustained metabolic benefits in obese mice

Encouraged by the acute results, we evaluated whether AzaA8 could provide longer-term metabolic benefits in animals with high-fat diet-induced obesity (DIO). Male C57BL/6N mice were fed a 60% fat diet from 9 weeks of age until they became obese (body weight ≥40 g at ~15–21-week old; see Fig. 4a), then allocated to treatment groups (n = 7–8 per group) for a 4-week intervention. AzaA8 was administered at 0.4 mg kg⁻¹ day⁻¹ (split into twice-daily intraperitoneal doses), and its effects were compared to daily liraglutide (0.4 mg kg⁻¹, i.p., twice daily as described above), intermittent semaglutide (0.12 mg kg⁻¹, i.p., every 3 days), or saline vehicle.

AzaA8-treated mice exhibited a significant reduction in body weight over the 29-day treatment, losing on average ~16% of their starting weight (Fig. 4b, c). This degree of weight loss was comparable to that achieved by liraglutide (~21.5%) at the same daily dose (~0.4 mg kg⁻¹) and was slightly greater than the weight reduction observed with the unlipidated semaglutide peptide (~9.5%, Aib8, 0.4 mg kg⁻¹ day⁻¹) (Supplementary Fig. 2). Semaglutide (administered at 0.12 mg kg⁻¹ every 3 days) also induced substantial weight loss (~19.8% in this model), as expected¹⁴. Thus, despite lacking the fatty-acid prolongation motif of semaglutide, AzaA8 delivered robust metabolic efficacy when given daily.

Glucose control was likewise improved by AzaA8. At the end of the 4-week treatment, mice were subjected to an oGTT. AzaA8 significantly blunted the glucose rise compared to vehicle-treated DIO mice (Fig. 4d). The glucose excursion in the AzaA8 group was similar to that in the semaglutide-treated group and markedly lower than in the vehicle group. Interestingly, liraglutide-treated mice did not show a significant improvement in oGTT glucose levels in this experiment, despite their weight loss. This contrasts with the typical profile of liraglutide and may relate to the specific dosing schedule or model variability^15,16.

To further characterize metabolic changes, we measured key hormones and biomarkers at the end of the study. Fasting serum leptin (Fig. 4e) and insulin (Fig. 4f) levels were both significantly reduced in the AzaA8 group relative to vehicle, consistent with a loss of adipose mass and improved insulin sensitivity. Meanwhile, fasting glucagon (Fig. 4g) and peptide tyrosine tyrosine (PYY) levels (Fig. 4h) were unchanged, and glycated hemoglobin (HbA1c) remained low (~4.5– 4.9%) and similar across all groups (Fig. 4i and Supplementary Table 7). The reduction in leptin in AzaA8-treated mice likely reflects decreased fat mass, as suggested by direct measurements of adipose tissue. AzaA8 significantly reduced the weight of subcutaneous inguinal white adipose tissue (iWAT) compared to vehicle (1220.8 ± 232.9 mg vs 1993.8 ± 328.8 mg, Fig. 4j). This reduction was comparable to that seen with semaglutide (1206.1 ± 340.9 mg) and liraglutide (907.9 ± 279.4 mg). Importantly, AzaA8 had minimal effect on the mass of interscapular brown adipose tissue (iBAT) (Fig. 4k) and did not cause loss of lean tissue such as skeletal muscle (quadriceps) (Fig. 4l) or significant liver weight change (Fig. 4m). Preservation of iBAT and muscle, alongside preferential loss of white fat, is a favorable outcome, suggesting that weight loss occurred mainly through depletion of excess white adipose tissue (Supplementary Fig. 3). Finally, to assess safety, we profiled inflammatory cytokines in the serum. We observed no significant changes in IL-6, IL-1β, TNF-α, or other cytokines in AzaA8-treated mice versus controls (Supplementary Fig. 4), indicating that the 4-week treatment did not provoke an adverse inflammatory response.

To test the clinically relevant subcutaneous route of administration (Fig. 5a), we performed a parallel four-week study in diet-induced obese mice. Daily injections of AzaA8 (0.4 mg kg⁻¹, s.c., daily), stabilized body weight relative to vehicle, preventing further gain, while semaglutide (0.04 mg kg⁻¹, s.c., daily) produced modest weight loss (Fig. 5b, c). Food intake was slightly reduced in the AzaA8 group during the latter half of the treatment period (Fig. 5d and Supplementary Table 8).

Glycemic control outcomes mirrored those of the i.p. study. At the end of 4 weeks, AzaA8-treated mice exhibited significantly improved glucose tolerance in the oGTT, with glucose levels during the test being the lowest among the groups (Fig. 5e). The total glucose AUC was significantly reduced by AzaA8 (764.7±81.5 mmol l⁻¹ min⁻¹) compared to vehicle (1742.5±104.9 mmol l⁻¹ min⁻¹) and even somewhat lower than in the semaglutide group (1195.4±89.2 mmol l⁻¹ min⁻¹) (Fig. 5f). Daily monitoring of non-fasting blood glucose provided additional insight: when measured 2 h after dosing, AzaA8-treated mice showed acutely lower glucose levels relative to vehicle (Supplementary Fig. 5a), whereas measurements taken just before each dose (pre-dose baseline glucose levels) showed no significant differences between groups (Supplementary Fig. 5b). This indicates that AzaA8’s glucose-lowering effect was closely tied to its dosing/exposure and that the HFD-fed mice in this timeframe did not develop overt fasting hyperglycemia. Consistent with that the above, HbA1c remained in the normal range (~3.6–3.8%) for all groups and did not differ significantly (Supplementary Table 9), reflecting that this DIO model, while obese and insulin-resistant, had not progressed to frank diabetes within the study duration. Overall, these chronic studies demonstrate that AzaA8 delivers sustained metabolic benefits, including weight reduction, preferential fat loss, improved glucose tolerance, and favorable hormonal changes, without triggering inflammation. The efficacy of AzaA8 across both intraperitoneal and subcutaneous routes of adminstration highlights the robustness of its therapeutic potential.

Discussion

Peptidomimetics with minimal backbone modifications can retain efficacy of native function while gaining important pharmacological advantages. In this study, we demonstrate that a single α-aza-amino acid substitution at a critical proteolytic site endows a GLP-1 analogue with markedly improved stability and in vivo efficacy. Replacing Ala8 with aza-alanine “locks” the N-terminus against DPP-4 attack, extending the peptide’s half-life and enabling sustained receptor agonism in vivo. This design leaves the peptide’s sequence and structure virtually unchanged aside from one atom, thus minimizing disruption of GLP-1R binding and likely reducing immunogenic risk^17,18.

Our results show that AzaA8 performs on par with established and clinically approved GLP-1RAs in animal models: it improved glucose tolerance acutely and produced significant weight loss or weight stabilization with chronic dosing, comparable to the effects of liraglutide and semaglutide. Notably, AzaA8 achieved these benefits without the need for fatty acid derivatization or other extensive modifications beyond the aza-substitution. The differences observed between our intraperitoneal and subcutaneous dosing studies (e.g. a greater magnitude of weight loss in the i.p. regimen) likely reflect differences in pharmacokinetics (exposure levels) and dosing frequency, as well as the comparator doses used. Nonetheless, both studies consistently highlight AzaA8’s potent metabolic effects and support its efficacy across different administration routes.

Our findings align with a recent report by Dinsmore et al.¹⁹, who independently identified that an aza-Ala8 substitution in GLP-1 analogues confers resistance to proteolysis and prolongs peptide half-life in vitro while retaining agonist activity. While Dinsmore et al. focused on in vitro stability and receptor assays, our work extends the evidence to in vivo efficacy, showing that an aza-GLP-1 analogue can deliver meaningful therapeutic benefits in animal models of diabetes and obesity. The convergence of these results from separate groups strengthens the case for aza-amino acids as valuable tools in peptide drug design. Collectively, these findings position aza-substitution as a rational and intrinsically efficient strategy to stabilize therapeutic peptides while preserving native function—marking a shift from empirical modification toward data-driven, atom-level peptide design that could simplify manufacturing and reduce development costs for durable peptide therapeutics.In summary, targeted aza modifications represent a promising approach to bolster peptide therapeutics. By blocking a primary proteolytic hotspot, we created a GLP-1 analogue that is both long-lasting and highly active, achieving significant anti-diabetic and anti-obesity effects in mice. This proteolytic-hotspot aza-subsitution strategy should be generalizable to other peptide hormones or therapeutics that suffer from rapid degradation, opening the door to more durable and effective peptide-based treatments for metabolic disease and beyond.

Methods

Reagents and chemicals

All chemicals and reagents were purchased from Fisher Scientific (Hampton, NH) or from Sigma-Aldrich (St. Louis, MO) unless otherwise noted. GLP-1 (7–37) was obtained from Tocris Bioscience™ (Minneapolis, MN), while semaglutide and exendin-4 were obtained from MedChemExpress LLC (Monmouth Junction, NJ).

Azapeptide analogues and peptide synthesis

Peptide controls and azapeptides were synthesized using standard reagents and methods on a Liberty Blue^™ 2.0 Microwave Peptide Synthesizer (CEM Corporation, Matthews, NC). Aza-amino acids and azaGLP-1 analogues were synthesized using our previously established azapeptide synthesis platform, employing benzotriazole-based building blocks and microwave-assisted SPPS¹³.

Structure verification and purity methods

Flash Chromatography and HPLC were used to purify all synthesized peptides and aza-analogues (purity level minimum >95%), and high-resolution mass spectrometry instrument (HRMS) was used to confirm sequence/structures (Agilent 6550 iFunnel QTOF LC/MS). Detailed support for characterization, structural confirmation and purities of the GLP-1 azapeptide analogues are provided in¹³.

In vitro protease stability: DPP-4 half-life assay

Stock solutions (5 mg ml⁻¹) of test articles and positive controls were made in sterile ddH₂O. Reactions were prepared by incubating 16 nmol l⁻¹ DPP-4 (D3446, Aldrich) at 37 °C, with a final test article concentration of 0.1 mmol l⁻¹ in sterile Tris HCl buffer (50 mmol l⁻¹, pH 7.4). The vials were kept in a 37 °C incubator for the duration of the experiment. Aliquots (30 μl) were taken at various timepoints (0, 1, 4, 8, 24 h) and added to vials which had been pre-filled with 20 μl of 0.05% TFA in water, and gently mixed. The samples were injected directly into the HPLC to monitor the recovery of intact peptide. The HPLC (Waters (Breeze QS System) was equipped with a 1525 binary pump, and a Phenomenex kinetex 2.6 mm EVO C18 analytical column (100 Å 150 × 4.6mm). Chromatography was performed at ambient temperature with flow rate of 0.7 ml min⁻¹ with linear gradient from water (0.05% TFA): MeOH (0.05% TFA) [95:5] to water (0.05% TFA): MeOH (0.05% TFA) [5:95] over 15 min and resolved peaks were detected by a 2998 photodiode array (PDA) detector at 215 nm. The peak area response ratio (PARR) was compared to the PARR at time 0 to determine the percentage of test article remaining at each time point. Half-lives of samples in DPP-4 were calculated using Microsoft Excel. The general methods were adapted from²⁰.

UPLC–MS/MS profiling of N-terminally modified GLP-1 analogues

Saline solutions (0.2 mg ml⁻¹) of each test compound were administered to male ICR mice (n = 2) via a single intravenous injection (1 mg kg⁻¹). Blood and urine samples (100–200 μL) were collected at 0, 15, and 60 min post-dose and immediately quenched in pre-chilled tubes containing 200–400 μL of acetonitrile:methanol (1:1, v/v) to precipitate proteins and halt proteolysis. Samples were vortexed and held on ice (~4 °C) until the final time point, then vortexed again and centrifuged at 18,200 ×g for 10 min. Supernatants were removed, diluted 1:1 with 2% acetonitrile in 0.1% formic acid (water), filtered, and analyzed by HPLC-MS/MS; when not injected immediately, extracts were maintained on dry ice (−78 °C) prior to analysis. Cleavage site mapping was carried out on a Waters ACQUITY UPLC–TQD with PDA detection (190–400 nm). Peptides were resolved on an ACQUITY UPLC HSS C₁₈ column (1.8 μm, 2.1 × 50 mm) with mobile phase A (water + 0.1% formic acid) and B (acetonitrile + 0.1% formic acid), 2 μl injections, 0.3 ml min⁻¹. MS was acquired by ESI⁺ with full-scan MS¹ (m/z 200–2,000) and product-ion MS/MS (m/z 100–2,000) using collision-energy ramps (e.g., 15–35 eV). Targeted transitions/precursors for each peptide (e.g., [M+3H]³⁺, [M+4H]⁴⁺) were subjected to production scans, and b/y ion series were assigned with a 10 ppm tolerance.

In vivo stability-murine pharmacokinetic (pK) studies

A plasma pK study (Pharmacology Discovery Services Taiwan, Ltd.) was performed in male ICR mice (n = 6) following single intravenous (IV) administration of test articles at 1 mg kg⁻¹. The semi-serial plasma samples were collected from three alternative animals at each time point at 0.05, 0.167, 0.5, 1, 2, 4, 6 and 8 h after IV administration. The body weight of each animal was recorded and exposure levels (ng ml⁻¹) of test articles in plasma samples determined by liquid chromatography–tandem mass spectrometry (LC-MS/MS²¹). The exposure levels (ng ml⁻¹) of test articles in plasma samples were determined by LC-MS/MS. Results were plotted for plasma concentrations versus time (mean ± SD). The fundamental pK parameters after IV administration were obtained from the NCA of the plasma data using WinNonlin (manual mode). The lower limit values of quantification (LLOQ) in plasma samples was 2 ng ml⁻¹, and the exposure levels below the 75% of LLOQ (1.5 ng ml⁻¹) were determined as below the limit of quantification (BLOQ). The fundamental pK parameters of test articles after IV (t_1/2, C₀, AUC_last, AUC_inf, AUC/D, AUC_extr, MRT, Vss and CL) administration were obtained from the non-compartmental analysis (NCA) of the plasma data using WinNonlin.

In vitro GLP-1 receptor potency measured by cAMP signaling

The in vitro potency assay (Novo Nordisk) relied on a reporter gene assay (CRE-luc) and was carried out in baby hamster kidney (BHK-21) cells overexpressing the hGLP-1 receptor. For these experiments, the compounds were tested in four technical pseudo-replicates. The combined results from three independent experiments were used to determine EC₅₀ values and data analyzed for significance at a 95% confidence interval.

Acute efficacy of AzaA8 in oral glucose tolerance test (oGTT) in lean mice

In vivo studies were conducted at Pharmaron (Beijing, China) in compliance with Institutional Animal Care and Use Committee (IACUC) protocols and Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC) guidelines. Based on the statistical power calculations, eight animals per group were used. Male C57BL/6N mice (21-week old, chow fed; SPF Beijing Laboratory Animal Technology Co., Ltd) were fasted for 6 h with free access to water before testing. At 2 h prior to oGTT, baseline blood glucose was measured, followed by a single subcutaneous injection of the test articles, AzaA8 (0.125 or 0.4 mg kg⁻¹), Aib8 (0.125 or 0.4 mg kg⁻¹), exendin-4 (0.125 mg kg⁻¹), or saline. Treatments were blinded to the investigators, and mice were monitored for clinical abnormalities (Supplemental Table 5). For the oGTT, body weights were recorded and at the start of the experiment mice were given 2 g kg⁻¹ of glucose (100 mg ml⁻¹) by oral gavage, and blood glucose was monitored at 0, 15, 30, 60, 90, and 120 min using an Accu-Chek Guide glucometer from tail blood.

Chronic in vivo efficacy in obese mice

Chronic efficacy studies were conducted at Pharmaron (Beijing, China) under IACUC- and AAALAC-approved protocols, designed to parallel the acute study. Male C57BL/6N mice (SPF Beijing Laboratory Animal Technology Co., Ltd) were placed on a high-fat diet (HFD; D12492, Research Diets, Inc.) at 7 weeks of age and maintained for 15 weeks before allocation into weight-matched groups (n = 8). Mice were monitored for clinical abnormalities (Supplemental Table 5).

For subcutaneous dosing, animals (18–21 weeks old) received daily treatment for 4 weeks: AzaA8 (125 μg kg⁻¹ during week 1, then 0.4 mg kg⁻¹ in weeks 2–4; once daily in week 2, twice daily from week 3 onward), semaglutide (0.04 mg kg⁻¹ daily), or saline vehicle. Treatments were blinded to investigators, and mice were monitored daily for health. Body weight, food intake, and non-fasting blood glucose were measured twice weekly (2 h post-dose and immediately before dosing; Supplementary Fig. 5). At study end, mice underwent an oral glucose tolerance test (oGTT) following a 6 h fast. Blood glucose was recorded before dosing (−2 h), test compounds administered, and a 2 g kg⁻¹ glucose challenge delivered by oral gavage at time zero. Glucose levels were monitored at 0, 15, 30, 60, 90, and 120 min using a handheld glucometer. Mice were euthanized after 120 min, and blood was collected for HbA1c analysis.

For intraperitoneal dosing, studies were performed at the Feinstein Institutes (protocol #24–0467). Male C57BL/6N mice (Taconic Biosciences, USA) were fed HFD (D12492) from 9 weeks of age for ≥10 weeks before treatment initiation (≥15-week old, ≥40 g). Animals (n = 8 per group) received AzaA8 or liraglutide (0.4 mg kg⁻¹ twice daily, i.p.), semaglutide (0.12 mg kg⁻¹ every 3 days), or saline vehicle for 4 weeks. Body weight was recorded twice weekly. At study completion, mice underwent oGTT as above, with glucose monitored up to 120 min using a Vet GlucoGauge (Covetrus). Blood and serum were collected for HbA1c (Mouse Hemoglobin A1c (HbA1c) Kits (Catalog# 80310, Crystal Chemistry, Elk Grove Village, IL)), cytokine profiling (IFN-γ, IL-1β, IL-6, IL-10, IL-12p70, KC/GRO, TNF-α using an MSD kit (Proinflammatory Panel 1 Mouse Kit, 7 plex, K15048G-1, Meso Scale Diagnostics, LLC (MSD), Rockville, MD), and metabolic hormone measurements (insulin, leptin, glucagon, PYY; using an MSD kit (U-PLEX Custom Metabolic Group 1 Assays, K152ACM-1, MSD, Rockville, MD)). Tissue depots (iWAT, iBAT, quadriceps muscle, liver) were excised, weighed, and frozen for further analysis.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Figure 1. GLP-1/semaglutide engagement with GLP-1R and sequence schematics.

a) Structural overlay of the GLP-1: GLP-1R complex (ligand, red; receptor, pink, PDBid: 6X18) and the semaglutide: GLP-1R complex (ligand, blue; receptor, light blue, PDBid: 7KI0), aligned on the receptor. The boxed inset enlarges the peptide–receptor interface to illustrate representative hydrogen bonds, salt bridges, and hydrophobic contacts. b) Linear map of R34-GLP-1 showing only residues that contact GLP-1R (highlighted and numbered); non-contact positions are gray. Schematic of glucagon-like peptide-1 (GLP-1) and analogues. Sequence and descriptive (abbreviated names) information for c) GLP-1 and R34-GLP-1, d) unlipidated semaglutide, semaglutide, and liraglutide, e) and azaGLP-1-based analogues. Red dashed lines represent the DPP-4 cleavage site. The protein and peptide structures depicted in panels a and b were created with PyMOL Molecular Graphics System, v.3.0 (Schrödinger).

Figure 2. Cleavage hotspots and dominant whole blood fragments of GLP-1 analogues.

a) Cleavage hotspots of N-terminus-modified peptides identified by tandem MS experiments. b) Peptide bond-specific cleavage intensities are shown for Aib8, AzaA8 and AzaH7. The vertical axis denotes peptide bonds (i|i+1), with circle size and color indicating relative units (RU) of cleavage intensity: red, primary (RU 1.0); orange, secondary (RU 0.5); yellow, low (RU 0.25). All analogues display recurrent hotspots within the central FTSDVSSY…EGQAAKEFIAWLVR region, particularly around bonds 17, 19–21, 23–24, and 28–29. c-e) Dominant whole blood fragments for c) Aib8, d) AzaA8, and e) AzaH7, including both the dipeptide and remainder resultant of DPP-4 cleavage, which were consistently detected at very low levels. Collectively, these results show that aza- and Aib-based substitutions reduce N-terminal degradation, but cleavage converges on shared internal hotspots that yield consistent, high-intensity fragments across analogues. Relative units are normalized to the highest value observed for each sample.

Figure 3. Effect of acute administration of AzaA8 versus controls on oral glucose tolerance test (oGTT) in lean (chow-fed) C57BL/6N mice.

b) Blood glucose levels over time (min) during oGTT, and c) total average area under curve for oGTT blood glucose levels – in male 21-week-old, chow-fed lean mice (C57Bl/6N) after single subcutaneous injection (treatment −2 h) of GLP-1 analogues or vehicle (saline) prior to glucose bolus (t = 0), n = 8 animals per group. Data are presented as mean ± SEM. Statistical analysis in b was performed using two-way ANOVA followed by Dunnett’s multiple comparisons test. P-values are indicated in the graphs and ****p< 0.0001, comparing each test article group to vehicle. When p-value is not indicated the comparison was not significant. Statistical significance in c was performed by one-way ANOVA followed by Tukey’s multiple comparisons test (all pairwise). P-values are indicated in the graphs and ****p < 0.0001. Panel a created with BioRender.com.

Figure 4. Effect of chronic intraperitoneal (i.p.) administration of GLP-1 receptor agonists on body weight and glucose levels in high-fat diet-induced obesity (HF-DIO) model.

a) The model consists in high-fat diet (60%)-fed C57BL/6N male mice (at least 15-weeks-old, or ≥40 g) treated for 29 days, b) bi-weekly body weights and, c) total body weight loss for AzaA8 (0.4 mg kg⁻¹ day⁻¹, BID, every 12 h, liraglutide (0.4 mg kg⁻¹ day⁻¹, BID, every 12 h), semaglutide (0.12 mg kg⁻¹ 3-day⁻¹, QD) or vehicle (saline BID, every 12 h), d) blood glucose levels after oral glucose tolerance test (treatment day 29) in HF-DIO mice treated daily as indicated for n = 7–8 animals per group. Effect of chronic (i.p.) administration of lead GLP-1R agonist-AzaA8 vs. controls on blood levels of selected diabetes- and metabolism-related proteins in HF-DIO model. Fasting e) leptin, f) insulin, g) glucagon, h) total PYY, and i) HbA1c levels in in C57BL/6N DIO mice after 29 days of treatment with vehicle (daily), 0.12 mg kg⁻¹ 3-day⁻¹ semaglutide, 0.4mg kg⁻¹ day⁻¹ liraglutide or AzaA8 (n = 7–8 mice per group). Effect of chronic (i.p.) administration of GLP-1 receptor agonists on isolated tissue weights in HF-DIO model. j) iWAT, k) iBAT, l) quadriceps skeletal muscle, and m) liver wet masses in C57BL/6N DIO mice after 29 days of daily treatment with vehicle (daily), 0.12 mg kg⁻¹ 3-day⁻¹ semaglutide, 0.4 mg kg⁻¹ day⁻¹ liraglutide or AzaA8 (n = 7–8 mice per group). Data are presented as mean ± SEM unless otherwise indicated. For panels b and d, group differences were analyzed using two-way ANOVA followed by Dunnett’s multiple comparisons test (vehicle vs. test article groups). For panel c, one-way ANOVA followed by Tukey’s multiple comparisons test (all pairwise) was used. For panels e-i, data are shown as mean ± SD and analyzed by one-way ANOVA with Dunnett’s multiple comparisons test (vehicle vs. test article groups). For panels j-m, data are shown as mean ± SEM and analyzed by one-way ANOVA with Dunnett’s multiple comparisons test (vehicle vs. test article groups). P-values are indicated in the graphs and ****p < 0.0001, when p-value is not indicated the comparison was not significant. Panel a created with BioRender.com.

Figure 5. Effect of chronic subcutaneous (s.c.) administration of GLP-1 receptor agonists on body weight and glucose levels in HF-DIO model.

a) High-fat diet (60%)-fed C57BL/6N male mice treated for 29 days with s.c. administration, b) bi-weekly body weights, c) total body weight loss, d) food consumption for AzaA8 (0.4 mg kg⁻¹ day⁻¹), semaglutide (0.04 mg kg⁻¹ day⁻⁻¹) or vehicle (saline), e) blood glucose levels after oral glucose tolerance test (treatment day 29), and f) total glucose AUC in HF-DIO mice treated daily as indicated for n = 8 animals per group; Data are presented as mean ± SEM unless otherwise indicated. In panels b, d and e, group differences were analyzed using two-way ANOVA followed by Dunnett’s multiple comparisons test (vehicle vs. test article groups). In panels c and f, group differences were analyzed by one- way ANOVA followed by Tukey’s multiple comparisons test. P-values are indicated in the graphs and ****p < 0.0001, when p-value is not indicated the comparison was not significant. Panel a created with BioRender.com.

Data availability

Source data are provided with this paper.