Emily A.
O'Brien
,
Mohaddeseh
Abbasi
,
Jeffrey A.
Purslow
and
Brett
VanVeller
*
Department of Chemistry, Iowa State University, Ames, IA 50011, USA. E-mail: bvv@iastate.edu
First published on 20th August 2025
Amidines are a relatively unexplored isostere of the amide bond, offering unique electronic properties and hydrogen-bonding behavior. This study presents the first systematic investigation of amidines within folded β-sheet structures. Using CD, NMR, and aggregation assays, we find that amidines are well tolerated when acting as hydrogen-bond donors but disrupt β-sheet folding when serving as hydrogen-bond acceptors. This donor/acceptor asymmetry contrasts with the behavior of amidines in α-helices, where both roles are accommodated. Importantly, outward-facing amidines disrupt the edge-to-edge hydrogen bonding required for fibril formation, enabling the design of non-aggregating β-hairpin peptidomimetics without reliance on sequence charge. Spectroscopic analysis further reveals that amidines embedded in peptide backbones exist predominantly in their neutral, monoprotonated form even at acidic pH, prompting a reassessment of amidine basicity in structured biomolecules. These findings establish design principles for using amidines in stable, aggregation-resistant peptidomimetic scaffolds.
Amidines differ from conventional peptide bonds by a single atom swap of the carbonyl oxygen for a nitrogen,16–18 and have served as design elements in pharmaceutical compounds.19–25 Specifically, Boger and coworkers have proposed that the different protic states of the amidine are responsible for the ability of vancomycin derivatives containing amidines (so-called maxamycins) to target antibiotic resistant Gram-positive bacteria.26–30 Amidines are also present in a class of natural products known as ribosomally synthesized and post-translationally modified peptides (RiPPs), which implies that an evolutionary advantage may have driven the development of biosynthetic mechanisms for their incorporation.1–3 Additionally, amidines have been proposed as potential intermediates in the prebiotic formation of peptides.31–33 Despite these examples from nature, there is a dearth of studies investigating amidines within synthetic peptides,14,15,34 which likely derives from a historical lack of robust methods for their incorporation into peptides.
Our initial report of a general methods to site-selectively install amidines in place of traditional peptide bonds35 created the first opportunity to explore the impact of this amide-bond isostere in α-helical structure.14 We found that amidines displayed intuitive hydrogen-bonding interactions that responded to pH and were amenable to design. The greater basicity of the amidine nitrogen led to stronger hydrogen-bond acceptor interactions that increased the stability and extent of helical folding. Ultimately, amidines appeared to be one of the few isosteric mimics of amides that could stabilize helical structure as both a hydrogen-bond donor and an acceptor.18
This work marks the first systematic investigation of amidines within folded β-sheet structures (Fig. 1). Amidines are a relatively unexplored single-atom isostere of the amide bond, offering unique electronic properties and hydrogen-bonding behavior. Using CD, NMR, and aggregation assays, we find that amidines are well tolerated in β-sheets when acting as hydrogen-bond donors, but disrupt folding when serving as acceptors. This donor/acceptor asymmetry contrasts with their behavior in α-helices, where both roles are accommodated. Outward-facing amidines further prevent fibril formation by disrupting the edge-to-edge hydrogen bonds required for self-association, enabling the design of non-aggregating β-hairpin peptidomimetics without reliance on sequence charge. Spectroscopic analysis reveals that amidines embedded in peptide backbones exist predominantly in a neutral, monoprotonated state even at acidic pH. Taken together, these findings position amidines as minimal, modular elements for engineering stable, aggregation-resistant peptidomimetic scaffolds—closely approximating native amides in conformation and polarity, unlike N-methyl amides or thioamides, which respectively distort geometry or increase hydrophobicity.
![]() | ||
Fig. 1 Hydrogen-bonding interactions of amidines in folded peptide secondary structure and effect on β-fibril aggregation. |
Peptide | Amidine | sequencea | ||
---|---|---|---|---|
Orientation | Strand | Turn | Strand | |
a All peptides except 8 are capped at the N-terminus with Ac and have a C-terminal amide (–CONH2). b Head-to-tail cyclic peptide. | ||||
YKL | RYVEV | pG | OKILQ | |
1-Tyr(NH)2 | ‘Out’ | RY(NH)VEV | pG | OKILQ |
2-Val(NH)3 | ‘In’ | RYV(NH)EV | pG | OKILQ |
3-Glu(NH)4 | ‘Out’ | RYVE(NH)V | pG | OKILQ |
4-Gly(NH)7 | ‘Out’ | RYVEV | pG(NH) | OKILQ |
5-Orn(NH)8 | ‘In’ | RYVEV | pG | O(NH)KILQ |
6-Ile(NH)10 | ‘In’ | RYVEV | pG | OKI(NH)LQ |
7 | RYVEV | PG | OKILQ | |
8 |
![]() |
With the native YKL sequence serving as a starting control, we synthesized35 six amidinopeptide analogues, featuring an amidine at different positions along the strand, both distal and proximal to the turn (Table 1). The positions of the amidine were selected to represent equal opportunities for the amidine to either point ‘out’ of the β-hairpin and into solution or point ‘in’ to the seam of hydrogen bonds that knit the β-hairpin together. Finally, we synthesized 7, which was assumed to be completely unfolded,10,12 through substitution of the turn-stabilizing D-Pro6 residue with L-Pro6. Closure of the β-hairpin into a cyclic peptide through inclusion of another D-Pro-Gly motif provided 8, which was assumed to be completely folded (Table 1).10 These peptides served as additional controls to quantify the extent of folding (vide infra).
The chemical shift assignments allowed us to estimate the nature of the folded structure for each residue (Fig. 2C). The difference between the 13Cα and 13Cβ chemical shifts has been shown to be an accurate and sensitive indicator of secondary structure, and even more accurate than Hα chemical shifts when distinguishing an β-sheet from a random coil.38–40 Specifically, compared to a random coil, negative ΔδCα–ΔδCβ values are indicative of β-sheet structure (where positive ΔδCα–ΔδCβ values indicate α-helical structure).41–43
The same dichotomy that was observed for the CD results was observed for the secondary structure analysis by NMR (Fig. 2C). All peptides for which the amidine pointed ‘out’ of the proposed β-hairpin fold generally displayed negative ΔδCα–ΔδCβ values for the residues associated with the β-sheet motifs (i.e. residues 1 → 5 and 8 → 12). Alternatively, all peptides for which the amidine pointed ‘in’ to the β-hairpin fold generally showed variable positive and negative ΔδCα–ΔδCβ values along the sequence, indicative of greater disorder and a poorly folded β-sheet. Finally, peptides that displayed similar ΔδCα–ΔδCβ values for their D-Pro residues to that of the YKL control peptide, were interpreted to have similar hairpin turn structure.
Further evidence for the folded β-hairpin fold was provided by the through-space interactions (NOEs) between residues across the β-strands, which are indicated in Fig. 3.44 This analysis was only carried out for the ‘out’ peptides which showed well-behaved folding by CD and chemical shift assignment (Fig. 2). Key signatures in the 1H–1H NOE spectra are the correlation between the amide proton (NH) and the Cα–H proton signals of residues across the strand,45 in addition to through-space interactions between side-chain atoms that should reside on the same side of the β-sheet. Specifically, through-space NOE correlations for the NH of Orn8 to the NH of Val5 are a signature of the β-turn motif.44,45 Many of the correlations that we observed for the YKL control peptide,44 were observed for the ’out’ amidino-β-hairpins, lending further support to the folded structures proposed here. Specifically, through-space NOE correlations were observed for Tyr2 to Gln12 for 1-Tyr(NH) and 3-Glu(NH) peptides. These residues are obviously distal to one another along the sequence, but their spacial proximity is strong evidence of β-hairpin folding. Interestingly, we observed fewer cross strand interactions for peptide 4-Gly(NH)7 compared to the other two ’out’ amidino-β-hairpins, despite separate evidence suggesting it does adopt a robust β-hairpin (vide infra).
![]() | ||
Fig. 3 Observed and spectrally-resolved through space interactions of folded amidinopeptides 1-Tyr(NH)2, 3-Glu(NH)4, and 4-Gly(NH)7 derived from 1H–1H NOE experiments. |
The first method examined the chemical shifts of the Gly7 diastereotopic Hα protons (eqn (1)).10,12,46,47 Where ΔδGly7 was the chemical shift difference between the two Hα protons for the Gly7 residue in the peptides under consideration (1–6) relative to the same value (ΔδGly7(8)) for the cyclic control peptide (8), which was considered to be 100% folded. The closer eqn (1) was to unity, the closer amidino peptides 1–6 were considered to be folded into a β-hairpin (Table 2).
![]() | (1) |
![]() | (2) |
Peptide | Amidine | Eqn (1) | Eqn (2) | ||
---|---|---|---|---|---|
Val3 | Orn8 | Ile10 | |||
a Blank entries in the table correspond to peptides for which the amidine at those residues (Val3, Orn8, Ile10) affected the chemical shift of the Hα,14 making it irrelevant for analysis. Each folded % in the structure above the table corresponds to an X = NH at that position while keeping amides (X = O) at the other positions along the backbone, and is an average of the calculated values of fraction folded (eqn (1), Val3, Orn8, Ile10) for that peptide. | |||||
YKL | 0.85 | 0.59 | 0.84 | 0.49 | |
1-Tyr(NH)2 | ‘Out’ | 0.82 | 0.42 | 0.73 | 0.60 |
2-Val(NH)3 | ‘In’ | 0.30 | 0.02 | 0.01 | |
3-Glu(NH)4 | ‘Out’ | 0.96 | 0.72 | 0.93 | 0.36 |
4-Gly(NH)7 | ‘Out’ | 0.61 | 0.61 | 0.81 | 0.48 |
5-Orn(NH)8 | ‘In’ | 0.03 | 0.03 | 0.08 | |
6-Ile(NH)10 | ‘In’ | 0.16 | 0.03 | 0.19 |
The second method considered the chemical shift of the Hα protons for residues Val3, Orn8, and Ile11 to be diagnostic of the fraction folded.10,12,46,47 To assess the extent of folding, the Hα chemical shift for those residues in amidinopeptides (1–6) were related to those same chemical shifts in both the unfolded (7) and folded (8) control peptides (eqn (2)). Again, the closer that eqn (2) was to unity, the more peptides 1–6 were considered to be folded into a β-hairpin (Table 2).
The results from both of these analyses confirm the CD and NMR characterization data discussed above (Table 2). All peptides for which the amidine pointed ‘out’ of the β-sheet displayed folding values on par with the YKL control peptide, while all ‘in’ peptides display much lower degrees of folding relative to the control. It should be noted that eqn (1) can overestimate the fraction folded;10,48 however, the values for eqn (2) provide confirmatory metrics that collectively indicate that amidines are fully tolerated as isosteric replacements of amide bonds in β-sheets, provided the amidine points ‘out’ and away of the seam of hydrogen bonds necessary to stitch the β-strands together. Alternatively, the amidine at residues Val3, Orn8, and Ile10 alters the chemical shift of the Hα away from standard values used in the development of eqn (2).14 However, eqn (1) compensates for this limitation by examining Gly7 residues instead of the amidine-containing residue, lending confidence to the conclusions of this analysis.
In contrast, amidines are tolerated as hydrogen-bond acceptors in α-helices, where the helical curvature allows the extra N–H bond to project outward into solvent (Fig. S44). This structural flexibility helps accommodate the amidine without compromising the fold.14 Supporting this interpretation, previous studies have shown that α-helices are more tolerant to side-chain mutations than β-sheets due to their cylindrical geometry and greater conformational plasticity.50 Moreover, the hydrogen bonds in β-sheets are stronger, more linear and tightly packed than those in α-helices,51 which may make them less tolerant to perturbation by amide isosteres. Our findings reinforce these structural distinctions and offer new insight into secondary structure compatibility by probing with amidines as a minimal, backbone-focused perturbation.
While steric crowding from the additional NH group of the amidine likely contributes to destabilization when oriented internally, we cannot exclude the possibility that changes in the hydrogen-bond register or local ϕ/ψ angles also play a role. In particular, the presence of β-branched residues (Val3, Ile10) at these positions may amplify the conformational strain induced by the amidine. Future computational or mutational studies will be valuable in dissecting these effects in more detail.
At 20 °C, we observed a strong peak for the amidine in 1-Tyr(NH)2 due to the isotopic enrichment (Fig. 4A). When the 15N decoupling in the indirect dimension was removed, we observed the amidine peak split into a doublet with a 1JN–H coupling of 92 Hz (Fig. 4B). The splitting and coupling value are consistent with a singly protonated, neutral NH species.53–55 The nearby doublet for the Gly7 amide NH provided a convenient in situ verification of the single protonation, further validating this finding (Fig. 4A). At temperatures above 20 °C, we observed broadening of the amidine signal beyond detection, indicating the amidine is undergoing chemical exchange on the intermediate NMR timescale.
As the temperature was lowered to 10 °C, extensive line broadening of the amidine signal was observed, suggesting the presence of millisecond-timescale conformational transitions and/or the presence of alternate conformational states (Fig. 4C and D).56 To further probe these states, we performed a 15N Chemical Exchange Saturation Transfer (CEST) experiment (Fig. 4E).57–59 In CEST, a weak radiofrequency field (B1) is applied to monitor the magnetization transfer to sites undergoing conformational exchange. We specifically targeted the amidine region (101 to 104.5 ppm) and decreased the temperature to 5 °C to increase signal-to-noise. The global minimum (G1) in the CEST profile signals the major state. If the global minimum (G1) is undergoing exchange with alternate states, additional weaker minima (E1–2) equal to the number of detectable alternate states are observed.58 The results of the CEST experiment revealed that the major populated state of the amidine (G1) undergoes conformational exchange with two additional minor states (E1 and E2) (Fig. 4E). Given the dominance of the singly protonated NH species at elevated temperature, we speculate that one of these two minor states is the conformational isomer of the singly protonated amidine (Fig. 4F). The other minor state could be the transient doubly protonated amidinium (Fig. 4F), which is a reasonable proposition given the pH of 4 and basic site present on the amidine. While beyond the scope of this work, these findings compel further experiments to understand the ensemble of states that amidines appear to undergo on the slow-to-intermediate timescale. This observation alone is surprising because the amidine 15N points into bulk solvent, which typically renders these exchanges too fast to detect on the timescale probed by CEST experiments. Such hidden states can be critical for evaluating the mechanism and binding of drug (macro)molecules to their protein targets.
Collectively, these data indicate that the dominant species for the amidine along the β-strand is the neutral monoprotonated state, even at the acidic pH of 4. This result contrasts with traditional views of amidines being highly basic (pKaH ≈ 12), and generally associated with poor pharmacokinetic properties in small molecules.60–62 However, we have found that amidines within peptide backbones can be considerably less basic (pKaH 5–6),14 perhaps due to reduced solvation of the amidine when placed within a peptide backbone. Regardless of the explanation, it is clear that amidines are far more tunable in terms of their basicity compared to the related guanidine functional group in the side-chain of arginine. Spectroscopic evidence indicates that arginine side chains are invariably protonated and charged (pKa > 12) in proteins, even when buried within hydrophobic microenvironments.63 Amidines do not appear to follow this behavior, however, an evaluation of the physical properties of more structurally diverse amidines is warranted.
To test this hypothesis, we synthesized three more peptides (Table 3). The YKL model peptide sequence contains three positively charged residues (Arg1, Orn8, Lys9) and does not display significant aggregation over extended time periods at the concentrations (4 mM) employed above. This behavior indicates why it has become a valuable model sequence for examining β-sheet behavior.10–13 We mutated two of the positively charged residues to neutral residues (R1Q, O8T) to give 9, which has neutral net charge. We observed a characteristic CD signature of β-sheet structure for 9 (Fig. S39, 100 μM), that matched the YKL control and indicated that the removal of charged residues was not detrimental to folding of the β-hairpin. By monitoring the transmittance of a 1 mM solution of 9, we observed an induction period65 followed by gelation that signaled aggregation of the peptide (Table 3, these results were confirmed by dynamic light scattering (DLS) experiments, see Fig. S43).
As a control, to probe whether the sequence of 9 was itself responsible for aggregation, we synthesized 10, which featured an L-Pro6 in place of the D-Pro6 in 9. Because the D-Pro6 is necessary for β-hairpin formation,10,12 peptide 10 was observed to be a linear random coil in CD experiments (Fig. S39). Accordingly, we did not observe any change in transmittance of a solution of 10 over the same time frame and concentration as 9 (Table 3 and DLS Fig. S43), indicating that aggregation was due to the β-hairpin structure and edge-to-edge fibrilization of the β-sheets.
Finally, we synthesized analogue 11-Tyr(NH)2 with an amidine at Tyr2 pointing ‘out’ of the β-hairpin (Table 3). We observed characteristic signatures of β-hairpin structure in analogy to 9 by both CD, NMR secondary structure and NOE analysis (Fig. S37–S39) that allowed us to estimate that 11-Tyr(NH)2 was folded to the same extent as the other ‘out’ peptides under consideration (Table 2). In contrast to 9, however, no aggregation or change in transmittance was observed (and DLS measurements aligned with the monomeric control peptide 10, Fig. S43). These results support our hypothesis that amidines pointing ‘out’ of the fold and into solution prevent aggregation by disrupting the hydrogen-bonding interactions necessary for edge-to-edge aggregation.
Together, these findings demonstrate that a single-atom substitution—a strategically positioned amidine in the β-hairpin—can prevent aggregation without compromising fold stability. Outward-facing amidines disrupt the directional hydrogen-bonding network required for edge-to-edge association, enabling a minimalist and modular strategy to inhibit aggregation. Crucially, this approach decouples aggregation resistance from sequence charge, expanding the design space for folded peptidomimetics targeting protein–protein interactions.
Outward-facing amidines inhibit aggregation by blocking edge-to-edge hydrogen bonding, enabling the design of non-aggregating β-hairpins without introducing net charge. This charge-independent approach expands the scope of sequence-compatible strategies for stabilizing peptide folds.
Interestingly, amidines pointed inward destabilize β-sheet folding, mirroring the behavior of thioamides, which generally act as poor hydrogen-bond acceptors in sheet contexts.11,12 However, the effects of thioamides are complicated by sulfur's hydrophobicity,66–69 making amidines a cleaner probe of backbone hydrogen bonding.
Finally, although amidines are traditionally viewed as highly basic, this assumption arises largely from solvent-exposed benzamidine analogs.24,60 Our spectroscopic data show that amidines embedded within peptide backbones remain largely neutral, even at pH 4. This finding calls for a reassessment of amidine basicity in structured environments and highlights opportunities for exploring new substitution patterns in peptide and peptidomimetic scaffolds.
Experimental and spectroscopic details, CD and NMR spectra for characterization of conformational structure. See DOI: https://doi.org/10.1039/d5sc05902j.
This journal is © The Royal Society of Chemistry 2025 |