Potent and Protease Resistant Azapeptide Agonists of the GLP-1 and GIP Receptors

Abstract

The gut-derived peptide hormones glucagon-like peptide-1 (GLP-1) and glucose-dependent insulinotropic polypeptide (GIP) play important physiological roles including glucose homeostasis and appetite suppression. Stabilized agonists of the GLP-1 receptor (GLP-1R) and dual agonists of GLP-1R and GIP receptor (GIPR) for the management of type 2 diabetes and obesity have generated widespread enthusiasm and have become blockbuster drugs. These therapeutics are refractory to the action of dipeptidyl peptidase-4 (DPP4), that catalyzes rapid removal of the two N-terminal residues of the native peptides, in turn severely diminishing their activity profiles. Here we report that a single atom change from carbon to nitrogen in the backbone of the entire peptide makes them refractory to DPP4 action while still retaining full potency and efficacy at their respective receptors. This was accomplished by use of aza-amino acids, that are bioisosteric replacements for α-amino acids that perturb the structural backbone and local side chain conformations. Molecular dynamics simulations reveal that aza-amino acid can populate the same conformational space that GLP-1 adopts when bound to the GLP-1R. The insertion of an aza-amino acid at the second position from the N-terminus in semaglutide and in a dual agonist of GLP-1R and GIPR further demonstrates its capability as a viable alternative to current DPP4 resistance strategies while offering additional structural variation that may influence downstream signaling.

Graphical Abstract

INTRODUCTION

The twin epidemic of type 2 diabetes (T2D) and obesity, termed “diabesity”, is a global public health problem. Current projections suggest that 10% of the world population will suffer from T2D within the next two decades.¹ Obesity (as defined by body mass index ≥30) has also seen a significant increase, with 42% of the current US population classified as such, and another 30% deemed overweight according to the Centers for Disease Control (CDC). The pursuit of safe and effective treatments for T2D and obesity has prompted investigation of peptides secreted in the postprandial state.² These molecules include glucagon-like peptide-1 (GLP-1)³ and glucose-dependent insulinotropic polypeptide (GIP), which both regulate glucose homeostasis and related physiological functions.⁴ These two ‘incretin’ peptide hormones, through the agency of their complementary roles in satiety signaling and accompanying weight loss and glycemic control, are at the very nexus of relief from liabilities arising from T2D and obesity.⁴ The stimulation of insulin biosynthesis and secretion by pancreatic β-cells takes place in a glucose-dependent manner modulated through the agonism of the cognate receptors of GLP-1 (GLP-1R) and GIP (GIPR). In peripheral tissue, GLP-1 delays gastric emptying in the stomach and induces satiety through signaling via vagal afferents in the brain, and GIP enhances hippocampal progenitor proliferation and regulates free fatty acid levels.²

A key modulator of the activity of GLP-1 and GIP is dipeptidyl peptidase-4 (DPP4), an ubiquitous serine protease that catalyzes removal of the two N-terminal residues of GLP-1 (His-Ala) and GIP (Tyr-Ala) (Fig. 1).^5–9 The resulting truncated forms have severely diminished activity profiles at their respective receptors.⁶ DPP4 exists within the epithelial and endothelial cells of many tissues including the liver. Several approaches have been utilized to make GLP-1 and related peptides more resistant to proteolysis. The most useful of these efforts have been (i) the use of an α/β peptide scaffold that changes the backbone of the construct,¹⁰ (ii) incorporation of fluorinated amino acids at strategic positions,⁵ (iii) the modification of side chains with saccharides,¹¹ (iv) use of thioamides as the peptide linkages,¹² (v) the use of 2-aminoisobutyric acid (Aib) at position 8 (second from the N-terminus) and more recently, (vi) the introduction of a new platform that relies on the modification of the N-terminus via alkylation.⁶ Substitution of hexafluoroleucine at the P1 or P1’ position of GLP-1 results in partial protection from proteolysis.⁵ N-terminal acylation of GLP-1 and GIP renders them refractory to enzyme-catalyzed hydrolysis; however, this approach leads to a significant diminution of agonism at the cognate receptors.^{6, 13} We have previously shown that N-terminal alkylations can be a powerful method to simultaneously provide protease resistance and maintain full stimulative activity at both GLP-1R and GIPR.^6–8 The more commonly used solution is the use of Aib at the P1 position of the substrate that occupies the S1 site of the enzyme active site. Synthetic analogues of GLP-1, and unimolecular dual agonists that exhibit both GLP-1 and GIP activity in the clinic have lately relied on this single modification. Through the use of Aib at position 2 of the peptide, agonists of GLP-1R and GIPR have resulted in leading peptide based clinical compounds in management of T2D and obesity (Fig. 1).^{14, 15} With the burgeoning interest in this field, the arsenal of chemical strategies that equip agonists with DPP4 resistance, and simultaneously do not interfere with receptor activation must be expanded. In this work, the replacement of canonical α–amino acids with an aza-amino acid (Fig. 2), which substitutes a nitrogen atom in place of the C_α atom, was pursued as a means of simultaneously preserving full efficacy and potency, modulating structural characteristics of the ligand–receptor complex, and to prevent the peptide from being inactivated by the hydrolytic action of the frontline protease.

Figure 1.

Amino acid sequences of GLP-1, GIP, semaglutide, and a balanced dual agonist (DA) of both the GLP-1R and GIPR with unnatural (blue), lipidated (green), and conserved (yellow) residues highlighted. The site of DPP4 catalyzed hydrolysis of GIP and GLP-1 is shown in addition to the aza-amino acid modifications (underlined).

Figure 2.

Left: Representative aza-amino acid containing peptide highlighting key structural differences from L-amino acids. The dihedral angle ϕ is approximately ±90° dictated by the lone pair–lone pair repulsion between the two nitrogen (blue) atoms. In addition, the planar urea [N(R²)–C(=O)–NH] moiety enforces further structural constraints and influences the three-dimensional disposition of the R³ and R⁴ groups.

An aza-amino acid-containing peptide ('azapeptide') features a semicarbazide functionality that perturbs electronic structure, and both the backbone and side chain conformations relative to the parent peptide. In addition, there is an observed loss of configurational permanence effectuated by the change from a tetrahedral α-carbon to a trivalently substituted nitrogen. This dynamic chirality thus expands the three dimensional disposition of chemical functionality that the molecule can display.¹⁶ Aza-amino acids have strong conformational preferences accompanying an additional lone pair–lone pair repulsion, and the coplanar nature of the urea moiety imposed through extended conjugation (Fig. 2, left). Similar modifications have been used to create 1,3,4-benzotriazepin-2-one tetrapeptide mimicking receptor agonists with selective signaling pathway activation profiles, making them ‘biased’ agonists.^{17,18,19, 20} While Aib has a preference for α-helical structures, aza-amino acids induce β-turn conformations, as judged by computation,²¹ X-ray crystallography,^{22, 23} and NMR spectroscopy.¹⁶ Compounds that can occupy and help populate different regions of the available conformational space may trigger 'bias' in the signaling pathways associated with GPCR activation.^{24, 25} The conformational space preferred by aza-amino acids is different from L-amino acids, making them a compelling addition to the toolbox to prevent enzymatic recognition of peptide substrates. Azapeptides, therefore, represent a new structural motif that can both be useful in conferring protease protection and because of conformational preferences, may dictate pharmacological efficiency and outcome.

RESULTS AND DISCUSSION

The residues of interest in GLP-1 and GIP chosen for aza-amino acid replacement were those adjacent to the scissile bond in the ligand that are subject to enzyme catalyzed hydrolysis. Accordingly, alanine at position P1 (second from the N-terminus in the peptide), and glutamic acid at P1’ (third from the N-terminus) that are present in both GLP-1 and GIP were first targeted for modification. As the second position from the N-terminus of these peptides is the most common manipulation point for abrogating DPP4 action, we explored three different aza-amino acids at this site. Aza-alanine (AzaA) was chosen as the closest structural proxy to the canonical residue alanine, and the insertion of aza-glycine (AzaG) and aza-proline (AzaP) were explored as substitutions that yield more or less flexible structures, respectively. The glutamic acid at position three (P1’ as a substrate) of GLP-1 was replaced with aza-glutamic acid (AzaE) with the intent to preserve side chain:receptor interactions (Fig. 1). In addition, the glycine at position 10 of GLP-1 was replaced with AzaG to assess the influence of a turn-inducing residue on G-protein signaling. Modifications at this position have previously established that both helix- and turn-promoting residues affect this pathway.²⁵

Azapeptides were assembled using standard solid-phase Fmoc peptide synthesis with substitutions based on the side chain of interest as discussed vide supra. This first necessitated the synthesis of the individual aza-amino acids that were chemically assembled using literature procedures which were altered slightly for efficiency (see Supporting Information). Briefly, the use of selective nitrogen differentiation on protected hydrazine was employed and followed by addition of the carbonyl donor by either using a phosgene equivalent (for AzaA and AzaP), 4-nitrophenyl chloroformate (for AzaE), or disuccinimidyl carbonate (for AzaG) as the activating agent. Boc-methylhydrazine was used to generate Fmoc-AzaA using a slightly modified version of a reported procedure.²⁶ AzaG was obtained as the Fmoc-protected hydrazine, and AzaP has a well-documented synthetic route.²⁷ The Fmoc-protected AzaE was synthesized using a combination of procedures.^{28, 29} These compounds were reacted with carbonyl-donor moieties to in situ generate the activated Fmoc-aza amino acid, that was transferred directly to resin.

We designed our azapeptides for testing on the two incretin hormone receptors, the GLP-1R and the GIPR. In addition to the native ligands GLP-1 (for GLP-1R) and GIP (for GIPR), we also tested our strategy on semaglutide, a clinical drug that is active on the GLP-1R, and a previously described unimolecular dual agonist (NNC0090–2746 from Novo Nordisk, termed ‘DA’ in this manuscript) of both GLP-1R and GIPR (Fig. 1).³⁰

GLP-1 and GIP are both extremely selective and potent agonists of their cognate receptors, which they agonize at single digit pM concentrations (Table 1). Since the frontline protease, DPP4, removes the two N-terminal amino acid residues from each of the peptide ligands that results in severe diminution of activity (> 99.9%), we first investigated the introduction of aza-amino acids at position two from the N-terminus of both GLP-1 and GIP and the capacity of the resulting constructs for robust agonism. To obtain potencies of the peptide, we utilized a concentration responsive cell-based bioassay that measures the ligand’s ability to stimulate the receptor and activate adenylyl cyclase to result in cAMP production, that is linked to the biosynthesis of luciferase. For measuring GLP-1R agonism, human embryonic kidney-293 (HEK-293) cells stably transfected with GLP-1R and the CRE_6X-luciferase reporter were used. GIPR agonism was assessed with HEK-293T cells transiently transfected with plasmids coding for the GIPR, reporter, and β-galactosidase to account for transfection variability.

Table 1.

Activities of GLP-1, GIP and their analogues on receptors.

Peptidea	EC50 (pM)b ± SEM	pEC50b	n c
GLP-1R
GLP-1 (native)	1.6 ± 0.14	11.8	3
A8AzaA GLP-1	2.4 ± 0.38	11.6	3
A8AzaG GLP-1	28 ± 3.6	10.6	3
A8AzaP GLP-1	96 ± 27	10.0	3
E9AzaE GLP-1	1200 ± 250	8.9	3
G10AzaG GLP-1	22 ± 2.0	10.7	3
DA X2AzaA, X20A	0.8 ± 0.19	12.1	3
Semaglutide	3.0 ± 0.27	11.5	3
X2AzaA Semaglutide	11 ± 1.0	11.0	3
GIPR
GIP, Native	6.3 ± 0.08	11.2	3
A2AzaA GIP	37 ± 0.07	10.4	3
A2AzaG GIP	120 ± 0.07	9.9	4
DA X2AzaA, X20A	2.4 ± 0.38	11.6	3

^aIdentity of the peptides that were evaluated (Fig. 1) for agonism at GLP-1R and GIPR using the luciferase reporter system in HEK-293 cells.

^bEC₅₀ is the concentration of peptide required for half-maximal activity at the target receptor. pEC₅₀ = −log(EC₅₀) ± SEM of independent experiments.

^cNumber (n) of independent experiments.

We first examined receptor stimulative properties of A8AzaA GLP-1, and remarkably this variant was extremely potent at the GLP-1R (EC₅₀ = 2.4 pM; Fig. 3a and Table 1) and was essentially equipotent and equiefficacious as native GLP-1. This outcome was both unexpected and dramatic as azapeptides like to adopt β-turn conformations, and the co-crystal structure of several ligands and GLP-1 or GIP show that residues at this position when bound to the receptor usually populate the α-helical region (the green/red markers in Fig. 4a). Further structural perturbations of the Ala8 site of GLP-1 with more and less structurally restricted aza-residues, AzaG and AzaP, resulted in observations that were more in line with expectations based on known azapeptide conformational preferences. When Ala8 was substituted with AzaG in GLP-1, it suffered a 17-fold loss in potency (Fig. 3b and Table 1), and the more rigidifying A8AzaP variant led to a more significant 60-fold decrease in potency when compared to native GLP-1 (Fig. 3c). These data suggest that the aza-residue that was the closest surrogate in terms of both size and shape (Ala8AzaA) was well accommodated by the receptor while others with access to larger (Ala8AzaG) or smaller (Ala8AzaP) conformational space were less effective agonists at the GLP-1R.

Figure 3.

Concentration-response curves from cAMP luciferase reporter assays: (a) A8AzaA GLP-1, (b) A8AzaG GLP-1, (c) A8AzaP GLP-1, and (f) Aib2AzaA semaglutide in comparison to native GLP-1 at the GLP-1R. Assays were performed using HEK-293 C34L cells stably transfected with GLP-1R and CRE_6x-luciferase. Both peptides engender similar G-protein mediated signaling through the G⍺s pathway. (d) Stimulative activity at the GIPR of Ala2AzaA GIP and native GIP illustrates a minimal potency decrease (< 6-fold), with a slightly higher potency loss observed in (e) Ala2AzaG GIP. A dual agonist (DA) 'DA X2AzaA, X20A' with the Aib2AzaA and Aib20Ala modifications is an equipotent and equiefficacious agonist compared to native ligands at both GLP-1R (g) and GIPR (h). Errors represent ± SEM.

Figure 4.

(a–e) Average Ramachandran plots for residue 8 of GLP-1 and its variants with GLP-1R derived from 5 parallel sets of molecular dynamics simulations. Trajectories from 100 to 300 ns were used for analysis. Angles (ϕ, ψ) of the following were shown for reference: Ala8 from GLP-1 (red circle, PDB 6×18); Ala2 from GIP (green circle, PDB 7RA3); Aib2 from semaglutide of (red star, PDB 7KI0); Aib2 from tirzepatide (green stars, PDB 7RGP and 7RBT). (f) The GLP-1:GLP-1R complex with residue 8 adopting the α-helix conformation with (ϕ, ψ) around (−62°, −37°). (g) The GLP-1:GLP-1R complex with residue 8 adopting (ϕ, ψ) around (−139°, −18°). (h) The GLP-1^A8AzaA:GLP-1R complex with residue 8 adopting the (ϕ, ψ) values around (−124°, −4°). His7 and residue 8 are highlighted in cyan. GLP-1 is shown in blue cartoon; the variant GLP-1^A8AzaA is shown in orange cartoon; GLP-1R is shown in gray cartoon. The residues of GLP-1R with heavy atoms within 5 Å of heavy atoms of residues 7 and 8 are highlighted in black, where green labels indicate the residues of GLP-1R analyzed in Fig S8 and Table S13, and pink labels are used for the remaining residues.

We then interrogated the C-terminal side (P1’ residue) of the scissile bond (DPP4 cleavage) by replacing the glutamic acid residue (third from the N-terminus, E9) with AzaE. This residue has previously been inspected using alanine scanning (Glu9Ala variant, 30-fold loss in binding as judged by IC₅₀ values, but equivalent adenylyl cyclase activity), or substitution by tert-Leu or β-dimethyl-Asp (equivalent cAMP production as native GLP-1).³¹ These findings encouraged us to incorporate AzaE at the P1’ position (third residue from the N-terminus) of GLP-1.³² Unexpectedly, the E9AzaE GLP-1 analogue suffered a dramatic 750-fold potency loss. These data suggest that the E9 residue occupies conformational space that the aza-amino acid residue is unable to adopt. Given the unexpected and surprising result of being able to modify position 8 of GLP-1, we also explored substitution at position 10 (fourth from the N-terminus) that is the P2’ site of the substrate with respect to DPP4. In this instance, alanine scanning resulted in greater than 10³–fold diminished adenylyl cyclase activity, and Gellman and co-workers have shown that replacement with an ACPC-(R,R–X) residue results in a significant loss in receptor stimulation.²⁵ Upon examining the G10AzaG analogue using the luciferase assay, we found that it suffers a 14-fold decline in potency compared to native GLP-1. These data show the relative promiscuity of the second position from the N-terminus with regard to substitutions on GLP-1, and that it can also accommodate aza-amino acids to various degrees with a preference for AzaA (Table 1). As mentioned previously, it was somewhat surprising to find that aza-amino acids could be used as they have strong (ϕ, ψ) dihedral angle preferences dictated by the diacyl hydrazine and the urea moieties. This prompted us to computationally investigate what solution backbone structures are favored in Ac-Xaa-NMe and His-Xaa-Glu-NMe, with Xaa being Ala, Aib, AzaA, AzaG, or AzaP, and whether they can occupy conformational space as those populated by natural amino acids in the GLP-1:GLP-1R complex (PDBID: 6×18), GIP:GIPR complex (PDBID: 7RA3), Semaglutide:GLP-1R complex (PDBID: 7KI0), tirzepatide:GLP-1R complex (PDBID: 7RGP), and tirzepatide:GIPR complex (PDBID: 7RBT) (Tables S3 and S4, Supporting Information).

Molecular dynamics simulations of the construct Ac-Ala-NMe showed that the Ala residue in this minimal construct adopted multiple conformations, e.g., in the β-sheet, Polyproline II (PPII), and α_R regions (Table S3a). As Aib, AzaA, AzaG, and AzaP are achiral, the Ramachandran plots of these amino acids in the dipeptide simulations were center-symmetric (Tables S3b–S3e). For Aib, these conformations include a major population in the α_R conformation and a minor population in the PPII conformation (and accompanying mirror conformations; Table S3b). On the other hand, AzaA favors the conformation around (–130°, 10°) along with a minor conformation around (−130°, −175°) (and the mirror conformations; Table S3c). AzaG predominantly adopts an extended conformation around (−155°, −170°) and the mirror conformation. AzaP exhibits two main conformations: a shifted PPII-like conformation around (−65°, −175°) and a conformation around (−75°, 10°) (and the mirror conformations). The broad distribution of Ala nearly encompasses the distributions of Aib, AzaA, AzaG, and AzaP in the negative ϕ region.

The Ramachandran plots for Xaa in His-Xaa-Glu-NMe showed that the addition of neighboring amino acids appears to favor the positioning of Ala and Aib in the α-helix region (Table S4). The (ϕ, ψ) values for Ala and Aib in these simulations generally align with those observed in X-ray co-crystal structures (green and red markings in Table S4). It is evident that AzaA prefers (–130°, 10°) and AzaP (–75°, 10°). On the other hand, AzaG continues to prefer an extended conformation.

Co-crystal structures show that GLP-1 when bound to the GLP-1R is predominantly in an α-helical conformation. However, closer to the N-terminus of the ligand, there is a region that is unstructured, and we postulated this section may be able to accommodate aza-amino acids and still engage with the receptor productively. In order to interrogate whether aza-amino acids can adopt conformational space that GLP-1 occupies in the ligand:receptor complex, we undertook molecular dynamics simulations of the native (GLP-1:GLP-1R) and modified complexes that substituted aza-amino acids at position 8 (second from the N-terminus) in native GLP-1, i.e., GLP-1^A8AzaA, GLP-1^A8AzaG, or GLP-1^A8AzaP bound to GLP-1R. Ala8 of GLP-1 was found mostly in regions of α-helicity, but also populated stable conformational regions that the aza-amino acid readily adopts (Fig. 4). This discovery was surprising, as it was thought that Ala8 would be confined to strictly α-helical regions similar to the Aib8 containing GLP-1 observed in the simulation of the receptor:ligand complex (Fig. 4b). We further note that while the AzaA residue in GLP-1^A8AzaA:GLP-1R populated backbone dihedrals around (–130°, 0°), close to the dihedrals observed in Ala8 in GLP-1:GLP-1R (Fig. 4c vs. 4a), the AzaG in GLP-1^A8AzaG:GLP-1R populated backbone dihedrals around (–150°, 0°) and the AzaP in GLP-1^A8AzaP:GLP-1R populated backbone dihedrals around (–60°, 0°) (Fig. 4d and 4e).

We also examined the crucial interactions involving His7 and residue 8 (Res8) of the GLP-1 ligands with the GLP-1R to assess the influence of these substitutions.³³ These include the interactions of the terminal α-ammonium group (–NH₃⁺) of His7 with the side chain of Glu387(receptor),^{34, 35} the NH₃⁺ of His7 with the side chain of Glu364(receptor),³⁵ the side chain of His7 with the side chain of Trp306(receptor),³⁶ the side chain of Res8 with the side chain of Leu384(receptor),³⁴ and the side chain of Res8 with the side chain of Leu388(receptor) as shown in Fig. S9.³⁴

The A8AzaA variant had similar interactions of comparable strength as those of the native peptide and GLP-1^A8Aib with GLP-1R. On the other hand, analysis of the dynamics of the GLP-1^A8AzaG:GLP-1R complex revealed less frequent interaction between AzaG8 and Leu388(receptor). In the case of the GLP-1^A8AzaP:GLP-1R complex, interactions between His7 and Trp306(receptor) were relatively sparse (Table S13).

Examination of the cryoEM structures of the GIP:GIPR complex also reveals an α-helical region for GIP reminiscent of the GLP-1:GLP-1R case, and a familiar unstructured N-terminus.³⁷ This observation and the success of the aza-amino acid containing GLP-1 analogues encouraged us to make azapeptide derivatives of GIP. We first tested a A2AzaA variant of GIP and found it suffered a minor loss in potency (5.8-fold) while being equally efficacious. On the other hand, an A2AzaG derivative of GIP underwent a more significant decrease in potency (19-fold) (Fig. 3d,e). We conclude from these observations that the native GIP:GIPR complex is more sensitive and less tolerant to subtle changes in backbone structure than it was in the case of GLP-1:GLP-1R combination. We then shifted our attention to making aza-amino acid variants of semaglutide (Ozempic) that has changed the landscape of T2D management and weight loss therapies. Popular in the public imagination and extremely effective as a clinical compound, semaglutide contains Aib at position 2 and we replaced that with an AzaA as we had previously with GLP-1. Semaglutide also contains a linker attached to a C18 diacid on a Lys side chain. This new azapeptide variant of semaglutide was nearly as equipotent and as efficacious as the parent drug and native GLP-1 (Fig. 3f).

The extent of aza-amino acid tolerance was further explored to expand the template inventory and document the generalizability of our approach. Unimolecular dual agonists (DAs) have recently emerged as potent compounds that enable higher percentages of weight loss, and we used a previously described template, NNC0090–2746, termed “DA” in this paper.³⁰ It is also known that the binding mode of some dual and triple agonists are slightly distinct especially at the N-terminus of the peptide when in complex with the receptor. DA is an unimolecular dual agonist compound that has been shown to be balanced, in that it is equally potent at both the GLP-1R and GIPR as compared to the respective native ligands. We introduced two modifications in DA, first 2-aminoisobutyric acid (X) at position 2 was changed to AzaA, and the X residue at position 20 was replaced with alanine (X2AzaA, X20A). A dual substitution of X at positions 2 and 20 to alanine (X2A, X20A) with all other structural characteristics of DA was equipotent at GIPR and GLP-1R compared to native ligands (unpublished data), and therefore we envisioned that it could tolerate the aza-amino acid modification at position 2. Our findings show that this new dual agonist azapeptide maintains the precedented balance at GLP-1R and GIPR with essentially equivalent potencies to the native ligands (Fig. 3g,h). We note that while an A2AzaA modification of GIP resulted in a comparative decline in potency at GIPR, and the A8AzaA modification of GLP-1 demonstrated equal potency at GLP-1R, a mutation to this dual agonist resulted in no shift in receptor balance or potency loss compared to native ligand at either receptor. This is presumably due to the subtle structural differences in the way the dual agonist and GIP bind to the GIPR.

Aza-modified GLP-1R and GIPR agonists were further tested for their ability to be refractive to DPP4-catalyzed proteolysis and inactivation (Table 2). We used two complementary methods to demonstrate this. First, we carried out an LC ESI-MS experiment where the azapeptide was incubated with DPP4 in an aqueous buffer, and the reaction mixture examined after 18 hours to assess the extent of proteolysis. The A8AzaA GLP-1 analogue was found to be completely resistant to DPP4 action and remained intact, while native GLP-1 was almost quantitatively degraded with no detectable full-length peptide remaining (Fig. 6a). We then employed the luciferase-based cAMP cellular assay to assess whether the protease can inactivate the azapeptide ligands. This latter method relies on the observation that the truncated peptides suffer a ≥10³-fold loss in potency. Briefly, peptides were incubated overnight with DPP4 and then the reaction mixture directly applied to cells and the concentration-response assay conducted as previously. A shift in potency after such an experiment reveals the extent of cleavage. This method is sensitive enough to detect even 0.5% of the peptide remaining whereas the LC ESI-MS experiments are more sensitive only when a smaller fraction of the peptides (≤ 80%) is degraded. For instance, if one half of the peptide were degraded, a potency loss of only 2-fold would be seen (within error of cellular assays, but easily discernible in LC-MS) whereas a 99.5% effective cleavage would result in a ~500-fold decrease in potency that would be easily detected by the cellular assay, but not so by LC-MS experiments. GLP-1R agonists were first tested including the native, A8AzaA, A8AzaG, A8AzaP GLP-1, X2AzaA semaglutide, and DA X2AzaA X20A (Fig. 5a–d, g, h). GIPR agonism of native GIP and the most potent azapeptide variant A2AzaA GIP were tested (Fig. 5e,f). All azapeptide agonists incubated with DPP4 showed no significant difference from azapeptide agonists incubated in the vehicle prior to the cell assay. These data demonstrate that agonists containing aza-amino acids at the second position to the N-terminus are not degraded by DPP4.

Table 2.

Stability against DPP-4 catalyzed hydrolysis of native and aza-amino acid containing peptides^a.

	−DPP4			+DPP4
Peptide	pEC50 ± SEMb	EC50 (pM)b	n c	pEC50 ± SEMb	EC50 (pMb	n c	Fold-Shift (↓)d
GLP-1R
GLP-1, Native	12 ± 0.07	2.2	4	8.8 ± 0.04	1.5 × 103	4	680
A8AzaA GLP-1	11 ± 0.05	4.7	3	11 ± 0.10	4.6	3	0.98
A8AzaG GLP-1	11 ± 0.08	20	3	11 ± 0.06	20	3	1.0
A8AzaP GLP-1	9.8 ± 0.05	160	4	9.5 ± 0.10	310	4	1.9
G10AzaG GLP-1	10	7.9	1	8.1	7.4 × 103	1	940
Semaglutide	11 ± 0.07	19	3	11 ± 0.06	16	3	0.84
X2AzaA Semaglutide	10 ± 0.11	52	3	10 ± 0.06	42	3	0.81
X2AzaA X20A DA	13 ± 0.01	0.29	3	13 ± 0.09	0.3	3	1.0
GIPR
GIP	11 ± 0. 17	6.1	3	8.9 ± 0.11	2.4 × 103	3	390
A2AzaA GIP	9.7 ± 0.08	210	3	9.7 ± 0.19	210	3	1.0

^aPotencies determined using HEK-293 cells expressing GLP-1R or GIPR with the luciferase reporter system. Results are separated by the targeted receptor. Peptides were incubated at 37 °C overnight with and without DPP4 before their incubation with transfected cells.

^bEC₅₀ = [peptide] required for half maximal activity of the targeted receptor. pEC₅₀ = −log(EC₅₀) ± SEM of independent experiments where applicable.

^cNumber of independent experiments.

^dRatio = (EC₅₀ with DPP4)/(EC₅₀ without DPP4).

Figure 6.

(a) LC-MS analysis of Ala8AzaA GLP-1 and native GLP-1 following their incubation with DPP4. (b) Hydrolysis kinetics of Gly-Pro-pNa catalyzed by DPP4 that was incubated for 30 mins with either vehicle (PBS, pH 8.0, grey circle), GLP-1 (2 μM, black diamond), A8AzaA GLP-1 (2 μM blue hexagon), A8AzaP GLP-1 (2 μM, green square), or linagliptin (1 nM, a known competitive inhibitor, red triangle) prior to addition of the substrate.⁴⁰ Azapeptides do not seem to affect substrate processing to any significant extent even at 2 μM, suggesting that they do not bind or interact with the DPP4 active site as they are comparable to the buffer control. Errors represent ± SEM in (b).

Figure 5.

Concentration-response agonism curves at the GLP-1R and GIPR from cAMP luciferase reporter assay: (a) GLP-1, (b) Ala8AzaA GLP-1, (c) Ala8AzaG GLP-1, (d) Ala8AzaP GLP-1, (g) X2AzaA Semaglutide, and (h) dual agonist (DA) X2AzaA, X20Ala pre-incubated with DPP4 or vehicle overnight. Assays were performed using HEK-293 C34L cells stably transfected with GLP-1R and CRE_6x-luciferase. Concentration-response curves at the GIPR of (e) GIP and (f) Ala2AzaA GIP were generated under the same pre-incubation conditions, but with transient transfection of HEK-293T cells. Errors represent ± SEM.

Three plausible reasons could explain the refractory nature of the azapeptides towards DPP4. First, the degradation may be kinetically slow (≥ 18 hours); second, because of structural differences, it is possible that the scissile bond is improperly aligned in the enzyme active site; or finally, due to the conformational preferences of azapeptides, they may not bind the enzyme active site at all. Incubation of native GLP-1 with DPP4 followed by quenching and LC-MS analysis showed that GLP-1 has an approximate half-life of 30 seconds under these conditions (Fig. S1). There is precedent for compounds such as diprotin A to be hydrolyzed at a slower rate, but a more likely scenario was that the azapeptide was a competitive inhibitor or was incapable of occupying the enzyme active site. Peptides containing aza-amino acids at the P1 position of the substrate have been documented to resist degradation.³⁸ In addition, azapeptides (P1 position) have also been known to be noncovalent competitive inhibitors, such as in the case of the hepatitis C virus serine protease.³⁹ In order to address which of these two possibilities is operational in this context, we sought to interrogate whether our analogues act as competitive inhibitors, or are just plainly not recognized by DPP4.

In order to distinguish between these two possibilities, we tested the ability of the azapeptides to inhibit the cleavage (hydrolysis) of a known chromogenic substrate, Gly-Pro-p-nitroanilide (Gly-Pro-pNA). We measured the rate of Gly-Pro-pNa (added last) by p-nitroaniline production with prior incubation of DPP4 with either (i) native ligand (GLP-1), (ii) azapeptide variant (A8AzaA, A8AzaP), (iii) linagliptin, a known competitive inhibitor, or (iv) vehicle (Fig. 6b). No appreciable differences were observed between the aza-analogues, GLP-1 (which was presumed in its truncated form by the time of substrate addition), and vehicle. In contrast, linagliptin severely diminished enzyme action.

These data once again highlight the aversion of the azapeptide variants of GLP-1 and GIP to bind the DPP4 active site to any significant extent. DPP4 has a particular liking for proline at the second position (P1) and one may advance the idea that the aza-analogue A8AzaP, the closest in structure to proline, may have affinity to the active site. This idea is negated in the experiment shown in Fig. 6b as A8AzaP failed to show an inhibitory effect, and was similar to the vehicle (Fig. S2a–d). A single atom modification (O --> S) reported in GLP-1 by Petersson and co-workers also conferred stability against DPP4 and comparable cAMP production; although it required changing the N-terminal His to Phe in GLP-1 because of intramolecular cyclization.¹²

CONCLUSIONS

Analogues of GLP-1 and GIP containing the aza-alanine residue at the second position from the N-terminus are refractory to DPP4-catalyzed hydrolysis and inactivation. Simultaneously these azapeptides suffer minimal to no loss of potency or efficacy at their cognate receptors. Rapid renal clearance also affects longevity in vivo, and while backbone modifications do not redress this concern, the compounds described in this paper are compatible with albumin binding lipid modifications that allay this liability. The superb activity profiles exhibited by the X2AzaA analogues of semaglutide and a dual agonist of the GLP-1R/GIPR, both exhibiting improved renal clearance profiles,^{14, 30} further underscores the utility and potential of aza-amino acids to expand the universe of non-canonical amino acids employed in peptide therapeutics in the realm of T2D and obesity management. Moreover, these data help to illustrate an already-known feature of aza-amino acids, in that they resist enzymatic degradation, and our study expands this space to the previously unreported serine proteases, the largest class of hydrolases.³⁸ The utility of aza-amino acids may also show promise in exploring their relative pharmacological profile on Class B G-protein Coupled Receptors (GPCRs) compared to native ligands for class B GPCRs. Within this class, receptors and the downstream signaling they engender is dependent on subtle variations in the structural and conformational attributes of the agonist. This outcome is of clear and present interest as a means of modulating other signaling pathways such as β-arrestin 1 and 2 recruitment and concomitant receptor internalization. Signaling triggered by these proteins has differential effects and has been postulated to play a role in determining efficacy of therapeutic outcomes.^{41, 42} Studies have already shown that a 'bias' in cAMP signaling for GLP-1R agonism can be encouraged by single or minimal modification with structurally rigid residues,^{24, 25, 43} and therefore the aza-amino acids may serve as another avenue for all templates to not only provide protection against the frontline protease, namely DPP4, but also influence signaling pathways. Studies along these lines are underway in our laboratories.

Supplementary Material

Experimental procedures, peptide synthesis and analytical characterization, protease assays and kinetics, and cAMP luciferase reporter assays. Simulation protocols and analyses.

ABBREVIATIONS USED

GLP-1

Glucagon-like peptide-1

GIP

Glucose-dependent insulinotropic polypeptide

Aib, X

2-aminoisobutyric acid

T2D

type 2 diabetes

DA

Dual agonist

WHO

World Health Organization