Amidine isosteric modification tunes proteolytic stability and activity

Abstract

The therapeutic potential of peptides is severely limited by rapid metabolism mediated by proteases. Traditional stabilization strategies often compromise pharmacological profiles by disrupting native conformation or physicochemical properties. Here we show that backbone amidine substitution offers a minimal, site-specific modification that enhances metabolic stability. Using the pentapeptide Leu-enkephalin as a model, amidine replacement of metabolically labile amides attenuated or blocked proteolysis in a position-dependent manner, with one analog preserving and even enhancing G-protein signaling at the μ-opioid receptor while reducing β-arrestin2 recruitment. Amidines thus provide a modular strategy to rescue promising peptide leads limited by metabolic instability, with the potential of conferring protease resistance without the need for wholesale scaffold redesign.

Site-selective amidine substitution transforms labile peptide bonds into protease-resistant isosteres while preserving receptor signaling, providing a minimal backbone-editing strategy to improve peptide stability.

1. Introduction

Peptides are an increasingly attractive modality for translating natural products into therapeutics.^1–4 A central challenge in the development of bioactive peptides is achieving proteolytic stability, ensuring that peptides can exert their biological activity before enzymatic degradation in serum, plasma, or the gastrointestinal tract. Proteolytic stability is therefore a fundamental component of peptide drug discovery, on par with efficacy and toxicity data.^4–6

A variety of strategies have been reported to attenuate proteolysis (Fig. 1).⁷ These can include the incorporation of heterocycles⁸ or alteration of the side-chain stereochemistry in the form of d-amino acids or α,α-dialkyl amino acid substitution.^9,10 Alteration of the native peptide backbone in the form of aza-peptides,¹¹N-methylation (and peptoids),¹² and N-amino derivatives,¹³ or more substantive modification of the backbone in the form of β-amino acids,^14–16 have all been shown to block proteolysis. These backbone-targeting strategies ward off proteases through direct modification of the peptide backbone and are distinct from secondary structure-stabilizing approaches such as cyclization and stapling.^17,18

Fig. 1. Chemical approaches to attenuate proteolysis. Most traditional strategies disrupt native peptide conformation, while amidines more closely replicate amide behavior through minimal chemical change while conferring protease resistance.

One aspect of how these modifications reduce protease susceptibility is that they often disrupt native peptide conformation or hydrogen-bonding patterns that are also critical for target recognition. For example, d-amino acid substitution, while offering protease resistance, can destabilize local secondary structures or abrogate critical chiral interactions with biological targets, resulting in loss of activity.¹⁹ Incorporation of α,α-dialkyl amino acids, such as 2-aminoisobutyric acid (Aib), restricts backbone flexibility, favoring 3₁₀ helices over native folds and sometimes misaligning binding epitopes essential for biological function.^20,21 In other examples, β-amino acid mutations in tumor-associated antigens can diminish or prevent MHC I affinity and T cell activation.²² Likewise, backbone modifications like aza-peptides,^23,24N-methylation (and peptoids),^25–27N-amino,²⁸ all introduce different biases to the native conformation of the peptide, which in some cases can be exploited to impart active conformations, but can also alter structure and folding necessary for function.

Conceptually, the ideal strategy to confer proteolytic resistance would retain the native side chains, stereochemistry, and backbone conformation required to preserve the pharmacophore and biological activity.^7,29 Rather than globally modifying the peptide, could one rescue a lead compromised by proteolytic instability through a minimal, site-specific chemical change at the vulnerable position(s)?

Because proteases initiate hydrolysis by activating the carbonyl oxygen of the peptide bond, most existing backbone strategies leave a key point of vulnerability intact. Targeting the carbonyl oxygen itself offers a more direct means to interfere with protease recognition. Thioamides exemplify this strategy, having been shown to reduce proteolysis in proteins by altering the electronic and structural properties of the bond.^30–33 However, thioamides introduce notable changes to the bond length (C S vs. C O),³⁴ hydrogen-bonding preferences,^35,36 and substantially increase the overall lipophilicity of the peptide,^37–39 which can compromise peptide folding, target binding, and bioavailability.^40–45

Amidines offer a promising alternative to traditional amide surrogates. We recently demonstrated that amidines display similar conformational behavior to native amides in folded peptide structures and closely replicate the hydrogen-bonding behavior of native amides.^46,47 Contrary to the prevailing view that amidines are highly basic, positively charged entities, their pK_a values approach neutrality when embedded in peptide backbones.⁴⁶ In other work, Boger and coworkers showed that amidine incorporation into vancomycin analogs preserved critical hydrogen-bonding interactions and retained biological activity against vancomycin-resistant bacteria.⁴⁸ Despite these encouraging results, amidines have remained largely underexplored in peptides.^49–51 Only recently did we develop a general method for the site-selective installation of amidines into peptide backbones, enabling broader access to this functional group in drug-like molecules.^52–57

This work details the first designed application of amidines to replace the carbonyl oxygen of the amide bond with an NH group creating a backbone motif that proteases are not evolutionarily adapted to recognize with the potential to preserve native behavior.^46,47 More broadly, amidine substitution can provide a practical framework for stabilizing peptide leads with short half-lives: identify proteolytic sites, install amidines selectively at those positions, and tune any potency shifts through complementary modifications when needed. This approach should extend to other therapeutic peptides where thioamide or N-methyl substitutions have proven too disruptive.

2. Results and discussion

2.1. Amidines confer protease resistance

We hypothesized that amidines could simultaneously (1) confer resistance to proteolytic degradation and (2) preserve the molecular interactions necessary for peptide bioactivity. To test this hypothesis, we selected the endogenous pentapeptide Leu-enkephalin (Leu-enk, YGGFL; Fig. 2), a short-lived but pharmacologically attractive agonist of the δ-opioid receptor (DOR). Leu-enk exhibits 1–5-fold selectivity for DOR over the μ-opioid receptor (MOR) and >1000-fold selectivity over the κ-opioid receptor (KOR), a profile associated with reduced risk of respiratory depression and addiction, and potential therapeutic applications in chronic pain, inflammation, and cancer treatment.^58,59 However, its poor pharmacokinetic profile (t_1/2 ∼ 10 min in plasma)⁶⁰ severely limits clinical application. In particular, rapid proteolysis at the Tyr₁–Gly₂ bond by aminopeptidase N in plasma, leads to rapid degradation.⁶¹ Thus, Leu-enk represented an ideal system to evaluate whether amidine substitution could enhance proteolytic stability without disrupting pharmacological function.

Fig. 2. Proteolytic activity of aminopeptidase N with Leu-enkephalin and amidine derivatives Y1 and G3 with aminopeptidase N (porcine kidney, supplied by Calbiochem) at 37 °C in pH 7.4 DPBS. Fraction intact peptide determined by UPLC absorbance of intact peptide at 280 nm and absorbance of cleaved tyrosine. Time points collected in triplicate, displayed as an average. Error bars represent one standard deviation.

We performed the first amidine backbone mutagenesis scan of a peptide, placing the amidine at the first three backbone positions along Leu-enk (Y1, G2, G3, Fig. 2). Synthetic limitations of our method to install amidines⁵² prevented the installation of an amidine between Phe₄–Leu₅ due to 5-exo-trig cyclization of the amidine onto the C-terminus.⁵³

As anticipated, we observed rapid proteolysis of Leu-enk with aminopeptidase N (t_1/2 = 16 min) in buffer. In contrast, when the amide at the Tyr₁–Gly₂ cleavage site was replaced with an amidine in Y1, the peptide was effectively intact over the same time frame. Interestingly, when the amidine was placed one and two positions away from the primary cleavage site in G2 and G3, some protease resistance was still observed (Fig. 2). These results support the first part of our hypothesis that the properties of amidines would confer protease resistance at the site of their installation. We next turned to what was arguably the more important aspect of the hypothesis: (2) The peptide, rendered proteolytically stable by the amidine, must still recognize and interact with its target.

2.2. Biochemical activity of amidine mutants

To test the second aspect of our hypothesis, we employed a cell-based assay to evaluate G-protein-dependent signaling at three opioid receptors (Table 1). This assay measures the potency of compounds Y1 and G3 in activating receptor-mediated signaling. It relies on bioluminescence resonance energy transfer (BRET) between two subunits of the heterotrimeric G protein: the α subunit fused to Renilla luciferase (Rluc) and the β/γ subunits fused to GFP2. Upon addition of the Rluc substrate (compound 400a), energy transfer from Rluc to GFP2 induces a measurable GFP2 fluorescence signal. Ligand-induced receptor activation (e.g., by Leu-enk and amidine mutants) promotes G protein dissociation, which reduces BRET efficiency and GFP2 fluorescence. This approach is a well-established method for assessing the signaling potency of opioid receptor agonists and is widely used to characterize their potential for analgesic activity.⁶²

Pharmacological profiling of Leu-enkephalin and amidine derivatives Y1 and G3 at opioid receptors MOR, DOR, and KOR. A, B, C represent Gi1 activation and D, E, F represent β-arrestin2 recruitment. E_max is reported as a percentage of the signaling induced by BW373U86, DAMGO, or U50,488 at the respective OR. SEM = standard error of the mean.

Ligand	MOR Gi1 activation		DOR Gi1 activation		KOR Gi1 activation
	E max ± SEM (%)	pEC50 ± SEM	E max ± SEM (%)	pEC50 ± SEM	E max ± SEM (%)	pEC50 ± SEM
DAMGO	100 ± 2	8.04 ± 0.06
BW373U86			100 ± 2	9.67 ± 0.05
U50,488					101 ± 11	9.14 ± 0.05
Leu-enk	97 ± 5	7.26 ± 0.13	85 ± 2	9.01 ± 0.07	96 ± 10	5.93 ± 0.14
Y1	93 ± 7	5.98 ± 0.13	80 ± 2	7.25 ± 0.08	N.D.	N.D.
G2	89 ± 8	5.88 ± 0.16	79 ± 3	6.93 ± 0.09	N.D.	N.D.
G3	96 ± 3	7.82 ± 0.08	73 ± 3	6.81 ± 0.10	91 ± 16	5.56 ± 0.24

	MOR β-arrestin2 recruitment		DOR β-arrestin2 recruitment		KOR β-arrestin2 recruitment
	E max ± SEM (%)	pEC50 ± SEM	E max ± SEM (%)	pEC50 ± SEM	E max ± SEM (%)	pEC50 ± SEM
DAMGO	100 ± 2	6.45 ± 0.05
BW373U86			99 ± 2	8.55 ± 0.05
U50,488					99 ± 2	6.87 ± 0.05
Leu-enk	94 ± 2	6.01 ± 0.05	94 ± 2	8.11 ± 0.05	N.D.	N.D.
Y1	N.D.	N.D.	80 ± 7	6.23 ± 0.18	N.D.	N.D.
G2	N.D.	N.D.	87 ± 5	5.95 ± 0.11	N.D.	N.D.
G3	100 ± 3	6.16 ± 0.06	69 ± 7	5.72 ± 0.16	N.D.	N.D.

At the μ-opioid receptor (MOR), compound G3 exhibited enhanced Gi1 activation (pEC₅₀ = 7.82 ± 0.08) compared to the parent compound Leu-enk (7.26 ± 0.13) and similar potency to the selective MOR activator DAMGO (8.04 ± 0.06). However, Y1 and G2 demonstrated lower Gi1 activation relative to Leu-enk.

Leu-enk is known to display higher affinity at the δ-opioid receptor (DOR) than at MOR, and the Tyr₁ residue in the native Leu-enk is notoriously sensitive to modification.^63–67 Accordingly, we observed decreased potency of Y1 at DOR (7.25 ± 0.08) relative to Leu-enk and the selective DOR activator BW373U86 (9.67 ± 0.05). The G2 and G3 amidine analogues were even less potent. It should be noted, however, that potency is only one aspect of the pharmacological profile, where a decrease in potency can be offset by protease resistance that maintains the drug at a higher concentration over time. Finally, all compounds were evaluated at the κ-opioid receptor (KOR). Leu-enk shows weak potency at KOR (5.93 ± 0.14) relative to the selective KOR activator U50,488 (9.14 ± 0.05). Accordingly, the amidine derivatives showed similarly weak activation of Gi1.

Another key aspect of opioid receptor signaling is the recruitment of β-arrestin2, a process typically associated with adverse effects such as receptor desensitization, drug tolerance (via receptor internalization),^68–70 and respiratory suppression—though the latter remains under debate.^71–73 To assess whether the amidine modification influences this pathway, we conducted a complementary, well-established cell-based assay to measure β-arrestin2 recruitment at the three opioid receptors (Table 1).⁶² This assay uses a similar BRET format to the Gi1 activation experiment described above. In this case, the bioluminescent enzyme Rluc is fused to the opioid receptor, and the addition of the luciferin substrate (Coh) generates a baseline luminescent signal. Upon ligand binding, β-arrestin2, fused to YFP, is recruited to the receptor, enabling BRET and resulting in YFP fluorescence—which serves as a quantitative readout of arrestin recruitment.

This β-arrestin2 recruitment assay was performed for all three opioid receptors (Table 1) In general, the potency of ligand-induced recruitment of arrestin was at least 10-fold lower than Gi1 activation for all compounds tested. At MOR, both G3 and Leu-enk exhibited comparable arrestin recruitment, consistent with their similar Gi1 activation profiles. At DOR, a similar 10-fold lower activation was observed for β-arrestin2 recruitment relative to Gi1 activation. As anticipated, none of the compounds tested demonstrated measurable recruitment at KOR.

These findings suggest that the observed reduction in β-arrestin2 recruitment relative to Gi1 activation may contribute to improved in vivo pharmacological properties. In particular, low β-arrestin2 efficacy, especially when combined with G-protein signaling, is expected to minimize adverse effects while preserving therapeutic benefit. Overall, the amidine-substituted analogs at the 1st and 2nd positions of Leu-enk showed reduced activity at MOR and DOR, and abolished activity at KOR, in the Gi1 dissociation and β-arrestin2 recruitment assays. Notably, the analog at the 3rd position, G3, exhibited a distinct profile at MOR, with increased Gi1 activation and similar β-arrestin2 recruitment compared to Leu-enk. Collectively, these findings support the idea that the N-terminal residues of endogenous opioid peptides are crucial for their agonist activity at opioid receptors.

Finally, a notable feature of the amidine analogues seems to alter the signaling bias across receptors. Leu-enk has been identified as a G protein-biased agonist at MOR and a balanced agonist at DOR and KOR.⁷⁴ Although potency shifts occur, the relative differences between Gi1 activation and β-arrestin2 recruitment observed for Leu-enk at MOR and DOR are recapitulated by Y1 and G2. Compared to Leu-enk, G3 shows further enhanced Gi1 activity with similar β-arrestin2 levels, suggesting that G3 improves G protein bias activity at MOR.

While the amidine replaces only the carbonyl oxygen of a single amide bond with an NH, it appears that this minor change still significantly impacts Leu-enk activity, likely due to alterations in the conformational dynamics within the orthosteric binding pocket.

2.3. Structural basis for preserved biochemical activity

To assess whether introducing an amidine group into the Leu-enk backbone induced structural changes associated with altered biological activity, we performed 1 µs molecular dynamics (MD) simulations of the Leu-enk and the amidine analogues complexed with μ-opioid receptor. Starting structures were generated from the crystal structure of DAMGO bound to the receptor (PDB code: 8EFQ).⁷⁵ The amidine was modeled in its NH tautomer because it is the only tautomer observed spectroscopically in peptides.^46,47 Because isomerization of the NH group is slow on the MD timescale, both E- and Z-isomers of the amidine NH group were simulated separately for all three mutants, designated Y1_E/Y1_Z, G2_E/G2_Z, and G3_E/G3_Z, respectively (Fig. 3).

Fig. 3. (A) Overlay of the backbone conformation of representative structures calculated from the 1 µs MD simulations on the peptide–µ-opioid receptor–Gi complexes. Peptides are shown as solid cartoons, and the receptor from the simulation is shown as a transparent cartoon. (B) Overlay of the representative structures from the simulations showing side chains. (C) The representative structures from the Leu-enk (left) and G3_Z (right) simulations highlighting intramolecular hydrogen bonds (yellow dashed lines). (D) Time evolution of hydrogen-bonding interactions during the simulations (X = O for Leu-enk and G3_Z, X = NH for G3_E).

The representative structures from the 1 µs trajectories revealed distinct differences between the E- and Z-isomers of G3, the most potent amidine mutant overall (Fig. 3A and B). While the backbone and side-chain orientation of G3_E aligned closely with DAMGO and Leu-enk complexed with the receptor, the backbone of G3_Z adopted a distinct conformation (Fig. 3A). Notably, however, this deviation was localized to the flexible Gly–Gly segment, which lacks side chains, and the overall side-chain orientations of G3_Z within the binding pocket remained consistent with those of DAMGO and Leu-enk.

The higher activity of G3 relative to Leu-enk at the μ-opioid receptor may arise from additional intrapeptide hydrogen bonds that stabilize its conformation (Fig. 3C). In Leu-enk, the conformation is stabilized by hydrogen-bond donation from Leu₅ to Gly₃ and from Gly₃ to Tyr₁. These same interactions are preserved in G3_E, consistent with its similar backbone conformation (Fig. 3D). In contrast, the altered backbone geometry of G3_Z across the flexible Gly–Gly segment reshapes its hydrogen-bonding network (Fig. 3C and D). New interactions appear between Leu₅ and Gly₂, as well as from the Phe₄ NH of the amidine to the Tyr₁ carbonyl. A further nonnative hydrogen bond is observed between the NH of the amidine at Gly₃ and the Phe₄ carbonyl. Despite these altered backbone interactions, the side-chain orientations within the binding pocket remain largely unchanged, suggesting that these additional intrapeptide hydrogen bonds may stabilize the bound conformation relative to Leu-enk and thereby enhance potency.

The 1 µs trajectories also provide insights into the lower overall potency of Y1 and G2. Both Y1_E and Y1_Z adopted backbone conformations distinct from those of DAMGO and Leu-enk (Fig. 3A), with correspondingly poor alignment of side chains (Fig. S6). These nonnative conformations likely contribute to the reduced activity of Y1 (Table 1).

Intriguingly, both G2_E and G2_Z displayed backbone conformations (Fig. 3A) and side-chain orientations (Fig. S7) that aligned well with those of DAMGO and Leu-enk, despite G2 exhibiting reduced potency relative to Leu-enk across the receptors tested (Table 1). This discrepancy can be rationalized by examining the time evolution of hydrogen-bonding interactions within G2_E and G2_Z during the simulation (Fig. S7). The intrapeptide hydrogen bonds that stabilize the conformation of the G2 isomers were weaker and more transient than those observed for G3. As a result, the conformation of G2 is more dynamic, which may explain its lower overall potency at the receptors.

3. Conclusion

Site-selective replacement of an amide with an amidine in Leu-enkephalin can block or attenuate proteolysis in a position-dependent manner, while preserving the relative Gi1/β-arrestin signaling bias even when potency is reduced. Even an amidine positioned two residues from the primary cut site imparted measurable protection.

Amidines complement established backbone modifications by offering a subtle way to tune proteolysis and signaling. By swapping the amide carbonyl oxygen for an NH, amidines disrupt protease engagement at a critical recognition element while largely preserving amide-like geometry, hydrogen-bonding behavior, and backbone neutrality in peptides. Together, these features make amidines a comparatively minimal backbone modification that can enhance proteolytic stability without substantially disrupting the conformational and physicochemical features needed for function.

Looking ahead, amidine substitution offers a practical roadmap for peptide leads with short half-lives: (i) map cleavage sites; (ii) install amidines selectively at those positions; (iii) balance any potency shifts with orthogonal approaches (e.g., side-chain modification) when needed. Importantly, even modest reductions in potency may be offset by the increased proteolytic stability conferred by amidine substitution, which can prolong peptide lifetime and maintain effective concentration relative to the native sequence.

4. Experimental

All experimental details and characterization data are provided free of charge in the supplementary information (SI).

Author contributions

J. B. D., K. K. S., and B. V. conceptualize the work and designed experiments. J. B. D and R. I synthesized the peptides. J. B. D and C. P. carried out experiments to support Fig. 2. S. M. B. and T. C. designed and carried out experiments to support Table 1. S. D. and V. V. designed and carried out experiments to support Fig. 3.

Conflicts of interest

J. B. D., K. K. S., and B. V. have filed a patent related to this work.

Data availability

All experimental details and characterization data are provided free of charge in the supplementary information (SI). Supplementary information: experimental and spectroscopic details, CD and NMR spectra for characterization of conformational structure. See DOI: https://doi.org/10.1039/d6cb00138f.