H-Bond Surrogate-Stabilized Shortest Single-Turn α-Helices: sp² Constraints and Residue Preferences for the Highest α-Helicities

Abstract

Short α-helical sequences of proteins fail to maintain their native conformation when taken out of their protein context. Several covalent constraints have been designed, including the covalent H-bond surrogate (HBS)—where a peptide backbone i + 4 → i H-bond is replaced by a covalent surrogate—to nucleate α-helix in short sequences (>7 < 15 amino acids). But constraining the shortest sequences (four amino acids) into a single α-helical turn is still a significant challenge. Here, we introduce an HBS model that can be placed in unstructured tetrapeptides without excising any of its residues, and that biases them predominantly into remarkably stable single α-helical turns in varying solvents, pH values, and temperatures. Circular dichroism (CD), Fourier transform infrared (FT-IR) absorption, one-dimensional (1D)-NMR, two-dimensional (2D)-NMR spectral and computational analyses of the HBS-constrained tetrapeptide analogues reveal that (a) the number of sp² atoms in the HBS-constrained backbone influences their predominance and rigidity in the α-helical conformation; and (b) residue preferences at the unnatural HBS-constrained positions influence their α-helicities, with Moc[GFA]G-OMe (1a) showing the highest known α-helicity (θ_n→π*MRE ∼−25.3 × 10³ deg cm² dmol^–1 at 228 nm) for a single α-helical turn. Current findings benefit chemical biological applications desiring predictable access to single α-helical turns in tetrapeptides.

Introduction

α-Helical sequences of ≤15 amino acids are key components of protein structure and function.¹⁻¹¹ Single α-helical turns¹²⁻¹⁶ in shorter sequences (four to six amino acids) also influence several biological activities in protein active sites, molecular recognition, protein–DNA interactions, and protein folding.^11,17−23 Hence, synthesizing single α-helical turn mimics has immense potential for biological applications. The problem however is that α-helices composed of ≤15 amino acids lack sufficient enthalpy from contiguous weak i + 4 → i backbone hydrogen bonds or i + 4···i inter-side-chain interactions, to counter the significantly high backbone unfolding entropy.^14,24−26 This enthalpy–entropy imbalance increases inversely with decreasing chain length and is the highest in tetrapeptides where only one i + 4 → i H-bond is possible. Tetrapeptides hence typically exist as random coils. Various structural modifications have been designed to overcome this entropic disfavor and to create a proclivity for the α-helical conformation in short peptides. These include the use of helix-nucleating templates,^10,27−33 unnatural amino acids,^21,34−36 metal clamps,³⁷⁻⁴⁴ noncovalent⁴⁵⁻⁴⁸ and covalent^{4,6,7,12,49−55} side-chain linkers, and covalent hydrogen-bond surrogates (HBSs).^52,56−59 Among these, the HBS strategy presents the advantage of not perturbing the native side-chain molecular recognition surface, which is essential for eventual functional mimicry. But an efficient HBS model is yet to be designed for constraining tetrapeptides in single α-helical turns with predictable high helicities.

The HBS strategy involves replacement of the labile main-chain hydrogen bonds with a covalent surrogate⁵² (Figure 1). The single α-helical turn is a 13-membered ring with a main-chain i + 4 → i (N_i+4–H···O=C_i–N_i+1) H-bond and contains nine sp² atoms and three complete residues. We present the term “13₉3 structures” to denote them. Early HBS models^1,52,60 were 13₈2 structures, containing two complete residues and eight sp² hybridized atoms in the HBS-constrained 13-membered rings, and have been excellent for nucleating α-helices^56,61−67 in longer (>7 residues) sequences. However, the i + 1st residue is replaced by an ethyl group in these HBS models and only the i + 2nd and i + 3rd residues are retained in the HBS-constrained ring. Due to lack of molecular recognition from both the C^α-stereochemistry of the i + 1st residue and its side-chain functional group, these HBS models cannot efficiently mimic single α-helical turns. Later, HBS models that retained all three residues in the HBS-constrained 13-membered ring were introduced.^57,58,68 These contained six sp² hybridized atoms (termed 13₆3 structures) and constrained short (≤6 amino acid) peptides with moderate conformational homogeneity and modest α-helicities (typical circular dichroism (CD) θ_n→π*MRE values of <−15 × 10³ deg cm² dmol^–1), compared to the helicities observed in infinite α-helix models⁶⁹ (whose CD θ_n→π*MRE values are ∼−43 × 10³ deg cm² dmol^–1). Hence, more efficient HBS models are needed to bias shorter (four amino acid) peptides predominantly into single α-helical turns.

Figure 1.

ChemDraw diagrams of the (a) i + 4 → i H-bonded natural single α-helical-turn 13-membered ring, containing nine sp² atoms and three complete residues (13₉3). (b) Current HBS-constrained 13₇3 models constrained in single α-helical turns, containing all three residues. Preferences for Gly and Ala residues at the i + 1st, i + 3rd, and i + 4th positions have been determined. (c) Earlier HBS-constrained 13₈2, 13₆2 models, lacking the i + 1st residue. The sp² atoms are numbered in each model.

We hypothesized that once an HBS that retains all three residues is in place, increasing the number of sp² hybridized atoms (from six to seven) in the HBS model can reduce ring flexibility and lead preferably to adopt majorly the α-helix. We thus designed the 13₇3 structures, containing all three residues and 7 sp² atoms in the HBS-constrained 13-membered macrocycle. To this end, (i) we replace the labile N_i+4–H···O=C_i–N_i+1 motif in various tetrapeptides by the covalent 1,3-propanediamine surrogate (N_i+4–CH₂–CH₂–CH₂–N_i+1) and (ii) introduce the N_i+1-methyl-oxycarbonyl (N_i+1–Moc) group (Figures 1 and 2c). The N_i+1–Moc derivatization introduces the seventh sp² atom. CD, Fourier transform infrared (FT-IR) absorption, one-dimensional (1D)-NMR, two-dimensional (2D)-NMR spectral and computational analyses show that these 13₇3 analogues have a propensity to adopt remarkably stable α-helical conformations in varying solvents, pH values, and temperatures, unlike their corresponding 13₆3 models (without the N_i+1–Moc group), which are disordered.

Figure 2.

Comparative (θ × 10³ deg cm² dmol^–1 vs λ nm) plots of (a) 1a–c and (b) 1a, 1d, and 1e in acetonitrile (100 μM, 295 K). (c) Labeled stick diagram of the current 13₇3 HBS model. (d) Residue preferences for each unnatural position in the HBS-constrained α-helical turns.

We also note that any HBS constraint introduces unnatural environments at i + 1st, i + 3rd, and i + 4th positions. Our current 13₇3 HBS model, for example, introduces a tertiary carbamate at the N-terminus of the i + 1st residue and a tertiary amide between the i + 3rd and i + 4th residues. As a result, the residue preferences at each of these unnatural positions need not be the same as they are in a natural α-helix, where alanine (Ala) has the highest helix propensity and glycine (Gly) has the lowest (excluding proline).^15,16,70 Determining residue preferences at these positions in the current HBS environment is essential for designing single-turn α-helices with high, predictable helicities, for various chemical biological applications. Hence, we synthesize 13₇3 analogues with Gly and Ala variations at these positions. Here, Ala is representative of the stereochemical environment at C^α of all non-Gly residues (excluding proline). By comparing the α-helicities of these 13₇3 analogues with those of the corresponding 13₆3 analogues and acyclic tetrapeptides, we determine the optimum sequences that render single α-helical turns with the highest α-helicities (θ_n→π*MRE ∼−25.3 × 10³ deg cm² dmol^–1) reported thus far. Based on 2D-NMR and CD data, reasons for consistent differences between α-helicities of highly stable HBS-constrained single-turn α-helices and the natural infinite helix models are also explored. Our results reveal the criteria for achieving maximum helicity and have immediate applications for designing mimetics of biologically important single α-helical turns, which form the crucial recognition element in several protein–protein interactions.^8,9,18

Results and Discussion

The list of 13₇3 and 13₆3 macrocyclic analogues and their corresponding acyclic tetrapeptides (1a–e, 2a–c, and 3a–c, respectively), synthesized following our earlier reported solution phase protocol for accessing short 3₁₀-helical mimics,⁷¹ are tabulated in Table 1. Residues in parenthesis are entirely contained within the 13-membered ring, constrained by the current HBS model. Phe was chosen at the invariable i + 2nd position for easy visualization of the molecules on thin-layer chromatography (TLC) and hence their facile purification. The AGADIR⁷² predictor showed <0.4% helicities (Table S6, Supporting Information (SI)) for all of the chosen sequences.

Table 1. List of HBS-Constrained 13₇3 and 13₆3 Analogues and Corresponding Acyclic Peptides Synthesized^a.

HBS-constrained cyclic peptidomimetics
cpd.	1373	cpd.	1363	cpd.	linear peptide analogues
1a	X-[G1F2A3]G4-OMe	2a	H-[G1F2A3]G4-OMe	3a	Y-G1F2A3G4-OMe
1b	X-[A1F2A3]G4-OMe	2b	H-[A1F2A3]G4-OMe	3b	Y-A1F2A3G4-OMe
1c	X-[G1F2A3]A4-OMe	2c	H-[G1F2A3]A4-OMe	3c	Y-G1F2A3A4-OMe
1d	X-[G1F2G3]G4-OMe
1e	X-[A1F2G3]G4-OMe

^aX, methyloxycarbonyl; Y, tert-butyloxycarbonyl.

The far-UV CD spectra of 1a–e, 2a–c, and 3a–c were recorded first in the apolar solvent CH₃CN in which backbone conformations and excitonic transitions will be weakly perturbed by solvent H-bonding interactions (HBIs), unlike in water. Concentration-dependent CD spectral analyses (60–500 μM, Figures S24, S27, and S30, SI) revealed nonvariance of the spectral shape and mean residue ellipticities for the n → π* transitions (θ_n→π*MRE, which occurs around 227 ± 1 nm), indicating the nonassociative nature of these analogues (Figure 3).

Figure 3.

Comparative CD (θ × 10³ deg cm² dmol^–1 vs λ nm) plots (CH₃CN, 100 μM, 295 K) of HBS-constrained 13₇3 (1a–c) and 13₆3 (2a–c) analogues, along with those of their corresponding acyclic (3a–c) tetrapeptides. The tetrapeptide sequences are (a) GFAG, (b) AFAG, and (c) GFAA, respectively.

Excitonic coupling¹⁰ between electronic transitions of peptide chromophores yields trisignate CD patterns (positive maximum at ∼194 nm for the π → π*_parallel transition and negative maxima at ∼208 and ∼222 nm for the π → π*_{perpendicular} and n → π* transitions), only in analogues where the electric dipole transitions of the peptides have an absolute helical orientation between themselves.^2,10 Among the 13₇3 analogues (1a–e), (Figure 2a) the α-helical triple band is clearly seen in 1a–c. The θ_n→π*MRE and θ_π→π*MRE magnitudes are the highest in 1a and are comparable to those observed in well-known single α-helical turn models^73,74 constrained by side-chain–side-chain covalent linkers, which were studied in water. This indicates a high α-helix content in 1a despite the lack of H-bonding interactions (HBIs) for the peptide chromophores, with either the solvent or any extended helical sequence. All three θ_n→π*MRE and θ_π→π*MRE intensities in the Gly_i+1 → Ala_i+1 analogue 1b are diminished (∼62% of 1a), indicating the weakening in delocalization of the excitonic energy over the three interacting peptide chromophores, which alludes to the perturbation of at least one of the peptide planes away from α-helical organization (compared to 1a). This is expected due to the increased steric strain in 1b, where all three HBS-constrained residues are C^α substituted (unlike in 1a). In 1c, the Gly_i+4 → Ala_i+4 analogue of 1a, the θ_n→π*MRE and θ_π→π*MRE intensities are low and the n → π* band is broadened up to ∼218 nm.^10,27 The CD of 1d, the A_i+3 → G_i+3 analogue of 1a, reveals the absence of selective preference for right- or left-handed structures. The shallow positive maximum at ∼199 nm and negative maxima at ∼218 and ∼229 nm in 1e, the analogue with Gly next to all of the tertiary carbamates, revealed a mixture of non-α-helical-turn structures²⁸ (Figure 2b). Thus, in the backdrop of the 13₇3 constraint, the residue C^α-stereochemistries further influence the α-helicity (the extent to which the constrained α-helical fold mimics a natural α-helical fold) and Gly is the most preferred residue at i + 1st and i + 4th positions, unlike in natural α-helices (Figure 2d).

For reliable correlation of CD spectral features of the constrained homologues with changes in their relative chromophoric arrangements, their conformations need to be stable. The CD patterns and θ_n→π*MRE values (CH₃CN) of 1a–c (the analogues with positive maximum helical structures) remain remarkably unchanged with the increase in temperature.⁷⁵ The thermal ellipticity coefficient (Δθ/ΔK, change in θ per degree change in temperature, the slope of the linear fit of θ_n→π*MRE vs T) of 1a (Figure 4a) is comparable to that of the most stable natural single-helical-turn peptide model reported by Baldwin et al.⁶⁹ Δθ/ΔK of 1b is closer to zero (closest reported yet) (Table S7, SI), consistent with rigidification of the macrocycle due to steric crowding. The helices also do not unfold with the increase in pH (Figure 4c) or concentration of urea (Figure 4b), the latter of which is known to solvate the peptide groups better than water and hence otherwise cause unfolding.²⁶ These indicate nonperturbation of interpeptide orientations, owing to which the influence of exciton–solvent interactions on CD patterns of the homologues is clearly discernible in current helical models.

Figure 4.

Variance of θ_n→π*MRE in 1a and 1b with (a) temperature (CH₃CN) and (b) concentration of urea (H₂O) and in 1a–c with (c) pH (30 mM buffers). The thermal ellipticity coefficients (Δθ/ΔK) for 1a and 1b (CH₃CN) are presented in (a). In all cases, the peptide concentration is 100 μM.

The helix diagnostic triple bands of 1a–c are retained in water (Figure 5a–c), as in CH₃CN, with the n → π* bands blue-shifting and the π → π* bands red-shifting, consistent with similar observations in uncoupled, isolated peptides.²⁹⁻³¹ Remarkably though, the magnitudes of these helically coupled θ_n→π*MRE and θ_π→π*MRE are enhanced by ∼1.6–2.0-fold (Table S8, SI) in water, compared to CH₃CN. The θ_n→π*MRE values are similar in the hydroxy solvents trifluoroethanol (TFE) and water (each of which have different dipoles but comparable strengths of H-bonds⁷⁶) but are slightly enhanced further in pH 7 buffer (Figure 5d), where the peptides additionally interact with ionic phosphates, with 1a showing one of the highest known values for a single α-helical-turn model^{33,68,69,73,77,78} (Table S7, SI). The CD pattern and hence θ_n→π*MRE remain relatively unperturbed by significant (×10⁷-fold) variations in [H⁺] (Figure 5c). The extents of such enhancements are typically lower in the corresponding acyclic analogues 3a–c (Table S9, SI), where the transitions are randomly coupled. These results reveal that solvent interactions perturb both the excitonic transition energies and their delocalization, when the chromophores are chirally organized even in a single α-helical turn; in addition, the magnitudes of such perturbations are strongly influenced by solvent–chromophore H-bonding and ionic interactions (than by protonation) of the coupled chromophores. An important outcome of these results is that the reference ellipticities for infinite α-helix models should have solvent-dependent (and not universal singular) values.

Figure 5.

Solvent-dependent CD spectra of (a) 1a, (b) 1b, and (c) 1c (100 μM) in CH₃CN, TFE, water, and pH 7 (30 mM phosphate) buffer; (d) variance of θ_n→π*MRE of 1a–c with different solvents (gray, CH₃CN; black, TFE; red, water; blue, pH 7).

CD spectra of the corresponding 13₆3 analogues 2a–c (N_i+1 is sp³ hybridized) lacked any notable CD bands (Figure 3a–c), consistent with earlier observations of conformational heterogeneity even in the N_i+1-sulfonamide derivatives of such analogues.⁵⁷ This reveals that merely linking N_i+4 and N_i+1 of tetrapeptides in H-bonding proximity is insufficient for biasing of their backbone predominantly into single α-helical-turn conformations. But once they are constrained, the number of sp² atoms in the macrocycle strongly influences largely α-helical conformation, with the threshold number being 7, as in the 13₇3 structures 1a–c. Lack of helicity even in 2b shows that it is not sufficient that all residues in the HBS-constrained macrocyle are C^α-trisubstituted (non-Gly). Both the number of sp² atoms and residue preferences are complementary and are important for high α-helicities of HBS models. This explains the loss in helicity observed in earlier HBS helix models upon reduction of the ethylene HBS¹ to ethane HBS.^63,79 The acyclic analogues 3a–c showed random-coil CD patterns (Figure 3a–c) as predicted by AGADIR,⁷² highlighting the propensity of the current HBS model to orient unstructured tetrapeptides primarily into single α-helical turns.

FT-IR spectra (CH₃CN, 2 mM) showed amide I Ṽ_C=O bands between 1654 and 1657 cm^–1 for both the 13₇3 analogues 1a and 1b (Figure 6), consistent with the predominance of α-helical-turn organization between the peptides.^22,37 To elucidate the structural stability and to further confirm the presence of α-helical conformation, the FT-IR spectra of N-deuterated 1a and 1b were recorded, which also yielded the amide I bands at 1656 and 1657 cm^–1, respectively. Absence of significant shifts in the C=O stretch bands indicated the nonexchange of the single α-helical turn in 1a and 1b to nonhelical conformations.⁸⁰ The 1639 cm^–1 band for 1c revealed predominance of β-turn structures (C′_i+1–C_i+2^α–C_i+3^α–N_i+4 torsion angle = 31.0°, which are descriptors of β-turns).⁸¹⁻⁸⁴

Figure 6.

Comparison of amide I Ṽ_C=O regions (CH₃CN, 2 mM, 295 K) in the FT-IR spectra of 13₇3 analogues, 1c–a (a–c); their deuterated amide (N–D) analogues (d) 1b (D) and (e) 1a (D), and their respective 13₆3 analogues (H₂O, 2 mM, 295 K) 2c–a (f–h).

FT-IR spectra (H₂O, 2 mM) of the corresponding 13₆3 analogues 2a–c showed an amide I band Ṽ_C=O at 1637 cm^–1, indicating a heterogeneous mixture of random-coil conformations in them (Figure 6f–h). This is consistent with the CD data indicating that the predominance of the single α-helical turn is induced in random-coil turns upon introduction of the seventh sp² atom in the 13-membered rings. Note that the 13₆3 analogues being ammonium salts were insoluble in organic solvents but were generally soluble in H₂O. Such a comparison of amide I bands in different solvents for the 13₇3 and 13₆3 analogues is agreeable since the FT-IR C=O stretching frequencies in α-helical and random-coil conformations are little perturbed by changes in solvent polarities⁸⁵⁻⁸⁷ (Table S10, SI). As expected, the acyclic analogues 3a–c showed similar random-coil values for the amide I Ṽ_C=O (1636–1639 cm^–1), confirming the role of the current HBS model in inducing α-helicity in unstructured tetrapeptide sequences.

¹H NMR (CDCl₃/CD₃CN 2:3) spectra of the 13₇3 analogues showed three sets of spin systems in 1a and two sets in 1b and 1c of which one predominated, in each. From 2D total correlation spectroscopy (TOCSY), ¹³C–¹H heteronuclear single quantum coherence (HSQC) and ¹H–¹H rotating frame Overhauser effect spectroscopy (ROESY) spectral correlations (Figure 7a–c), and the relative intensities of the signal sets (see SI), the major/minor ratios were determined to be 77:15:8 (1a), 74:26 (1b), and 89:11 (1c). The presence of ROESY exchange peaks between protons of the major and minor species indicated that they are conformational isomers (Figures S7 and S20, SI).

Figure 7.

ChemDraw diagrams of the crucial ROE cross-peaks in the major conformers of (a) 1a (77%), (b) 1b (74%), and (c) 1c (89%) and the stick and ribbon representations of their corresponding energy-minimum structures, in (d)–(f) showing only the Hs exhibiting ROEs. The phenyl group of Phe_i+2 is removed for clarity. The rise (Å) in the helical turn is denoted by C_i+1^α···C_i+4^α distance.

The structures of the major conformers of 1a–c (Figure 7d–f) were calculated using their ROE distance restraints. The ROEs were classified as strong, medium, and weak, and their distance boundaries were assigned (Table S11, SI). The initial random structure was optimized in Automatic Topology Builder (ATB)³⁹ using the semiempirical quantum mechanics (QM) theory,⁴⁰ followed by energy minimization using the GROMOS 54A7 force field⁴¹ in a box of pre-equilibrated simple point charge (SPC) water molecules. In 1a and 1b, the HBS-constrained tripeptide backbones make a right-handed turn, with the propyl group supporting α-helical rise (from C_i+1^α to C′_i+4), of 5.1 and 5.2 Å, respectively, comparable to those in ideal natural α-helices (5.4 Å). In 1c, the backbone twists away from a right-handed turn at residues i + 3 and i + 4. The N–Moc carbamate plane is similarly obtusely inclined from the propagating helix axis in 1a–c to avoid 1,2 strain between the Moc group and C_i+1^α substituents. This results in an upflip of the sp² hybridized N_i+1 and hence an unnatural positive value for φ_i+1, a common feature in these analogues (Table 2). The rest of the backbone φ, ψ dihedral angles are negative in 1a–c and consistent with those in the classical α-helical turn^43,44,88,89 (with 1a showing the least root-mean-square deviation (RMSD), Table 2 and Figure 8) and with the crystal structure of the olefin HBS model.⁵⁹ The s-cis and s-trans rotamers of the methyl carbamate group at the N_i+1 nitrogen are energetically indistinguishable. There are no ROE cross-peaks for the methyl group of methyl carbamate either, indicating that the carbamate group is remotely disposed from the α-helical turn. Hence, N_i+1 nitrogen has the potential to anchor various carbamates and acyl groups without affecting the 13₇3 HBS-constrained α-helical-turn conformations.

Table 2. List of (φ, ψ) Backbone Dihedral Angles and Backbone RMSD between Helical Folds in 1a–c and Their Natural Sequences.

helix models	(Φ, ψ)i+1	(Φ, ψ)i+2	(Φ, ψ)i+3	(Φ, ψ)i+4	RMSD (Å)
α-helixa	–65, −40	–65, −40	–65, −40		0
[xQV]Ab		113, −46	–71, −47	–55, −45
1a	83, −122	–88, −39	–72, −49	–75, −42	0.41
1b	47, −99	–102, −5	–107, −54	–78, −34	0.64
1c	65, −93	–87, −26	–92, −2	63, 36	1.27

^aFlanagan et al.⁴⁴

^bArora et al.⁵⁹

Figure 8.

C′_i=O_i···C′_i+1=O_i+1 dihedral angles representing interchromophore orientations and the C′_i···C′_i+1 distances are tabulated for 1a–c. Backbone superimposition of the GROMOS energy-minimized 1a–c (in green), with their corresponding natural α-helices (built with Discovery Studio) (in GBR)(top view).

The folds in 1a and 1b are similar, while that in 1c is significantly different, substantiated by comparable ¹H and ¹³C NMR chemical shifts (Figure 9 and Table S4) for 1a and 1b and not 1a and 1c. In both 1a and 1b, the crucial exocyclic φ_i+4 torsion is negative, as in a propagating helix. This is imperative for application of these HBS models for probable propagation of α-helicity in sequences appended at their C-terminal. However, in 1c, it is positive as in a nonpropagating turn and results in an oblique twist of the i + 3–i + 4 peptide plane. Imposition of negative values for φ_i+4 in computational models of 1c resulted in multiple unavoidable intramolecular van der Waals steric clashes involving C^βH₃ of Ala_i+4, indicating that propagation is probable at Gly_i+4 but is precluded by Ala_i+4.

Figure 9.

Tables of (a) ¹³C and (b) ¹H, chemical shifts (ppm) of 1a–c (CDCl₃/CD₃CN, 2:3), 5 mM, 295 K. Atomwise bar plots of corresponding Δδ ppm (δ_1a–δ_1b and δ_1a–δ_1c) values from (c) ¹³C and (d) ¹H NMR spectra.

The unnatural i + 3–i + 4 tertiary amide bond adopts the s-trans rotamer in all analogues, and the s-cis rotamer is sterically disallowed. In the cis rotamer, the C=O has to flip inward of the macrocycle. This is made impossible because the C^α-substituents of both i + 3rd and i + 4th residues infringe severely into each other’s hard-sphere van der Waals space in this rotamer. Alternatively, in the cis rotamer, the i + 4th residue backbone has to flip inward, which is all the more impossible. Note that the s-cis rotamer is reported to be disallowed even in the 13₆3 analogues, with a lesser ring strain.^57,58

The ROE-consistent GROMACS-derived minor conformers of 1a and 1b are due to the flipping of the i + 3–i + 4 peptide plane away from the helix axis, which makes it almost perpendicular to the other two peptide planes (Figure 10). This conformer is a kinetic minimum. 1a, which contains two highly flexible Gly residues (at i + 1st and i + 4th positions), shows an additional such (third) conformer (8%) in ¹H NMR, whose structure we are unable to characterize due to its low relative population and hence the lack of any corresponding spectral signals, especially in the 2D-NMR. It can at best be speculated that the third rotamer (and the minor conformer of 1c) is one of the kinetic minima like those reported by Broussy et al.,^57,58 where there is flipping of one or more of the peptide planes.

Figure 10.

Energy-minimized structures of minor conformers of (a) 1a and (b) 1b. (c) The list of (Φ, ψ) values from the energy-minimized structure of the minor conformers.

Interchromophore orientations⁹⁰ (Figure 11a) (defined by C′_i=O_i···C′_i+1=O_i+1 dihedral angle, C′_i···C′_i+1 distance, O_i···C′_i+n distance, O_i···C′_i+n=O_i+n angle, and C′_i=O_i···C′_i+n=O_i+n torsion, n = 1, 2; Table S12, SI) strongly influence CD patterns and hence are good markers for comparative analyses of NMR and CD data. Comparison of these parameters in 1a–c (Figure 11b–d) with those in the natural classical α-helical turn containing the corresponding sequences (built in Discovery Studio) revealed that for 1a the interpeptide torsions and angles are nearly comparable, consistent with its remarkably high CD θ_n→π*MRE values. However, the interpeptide separation increases by ∼0.6–0.7 Å more than in the natural α-helix, which is expected since the HBS motif occupies a larger volume compared to the natural H-bond that it replaces in the helix. Since excitonic couplings decrease by the inverse cube of the interchromophore distance,³⁴ this explains why θ_n→π*MRE for 1a is not the same as that for natural infinite α-helix models⁶⁹ (∼−43 × 10³ deg cm² dmol^–1) but is lower. Similarly, increased distances are expected in most HBS-constrained single-turn α-helix models, which explains their consistently lower θ_n→π*MRE values than natural α-helices. Considering the structural similarity between 1a and the natural α-helix, we present its θ_n→π*MRE (∼−25.3 × 10³ deg cm² dmol^–1 at pH 7) from current studies to be representative of the infinite α-helicity values for HBS-constrained single α-helical turns. The value decreases in weakly H-bonding solvents. In 1b, the intercarbonyl (C=O_i···C=O_i+1) dihedral angles are distorted for the i + 1 → i + 2 peptide pair (Figure 8). Their peptide separations are comparable, too, but still greater than those in 1a. Both interpeptide orientations and separations thus influence their decreased helicities. In 1c, the distances and orientations of both peptide pairs are significantly perturbed from canonical values, leading to poor helicities. Thanks to the rigidity enforced in these HBS-constrained structures, the influence of such subtle differences in interpeptide orientations due to variations in sequence, on CD bands, is clearly observable.

Figure 11.

(a) ChemDraw representation of interchromophoric orientation. Comparison of intercarbonyl (b) torsion (C′=O···C′=O), (c) angle (O···C′=O), and (d) distance (O···C′) between HBS-constrained peptides in 1a (black), 1b (red), and 1c (blue) with those in the classical α-helix (empty) built from corresponding sequences with Discovery Studio.

Note that unlike the N–Moc-protected 13₇3 analogues 1a–c (which are soluble in 40% CDCl₃ in a CD₃CN solvent mixture), their corresponding 13₆3 analogues 2a–c are ammonium salts and are generally soluble in water, in which their NMR spectra are recorded (Table S5, SI). Due to the difference in solvent conditions, the NMR chemical shifts⁹¹ and ³J Hz scalar coupling values⁹²⁻⁹⁴ of 1a–c and 2a–c are not comparable (Figure S23, SI). However, the conformational states of 13₆3 analogues similar to 2a–c have been extensively analyzed earlier⁵⁷ and were found to contain up to 10 different conformers, with no predominance of any ordered structures in the mixture. This is consistent with CD and FT-IR data of the current 13₆3 analogues, 2a–c. The predominance of α-helical conformations in the 13₇3 analogues 1a–c, on the other hand, based on their ¹H, ¹³C, 2D-NMR data, and molecular dynamics simulation analyses, thus further confirms that a threshold number of seven sp² atoms is essential for such conformational predominance in HBS-constrained peptidomimetics.

Conclusions

A covalent H-bond surrogate (HBS) model is presented for constraining unordered tetrapeptides in 13-membered-ring single-α-helical-turn conformations with high stability under varying conditions of temperatures, solvents, and pH values. In this HBS model, the n-propyl group replaces the i + 4 → i H-bond and the N–Moc group sp² hybridizes the N_i+1 atom. In the backdrop of the HBS-constrained 13-membered rings, strong influence of the number of sp² atoms is observed as to obtain predominantly single α-helical-turn conformation. Seven sp² atoms are essential, with six yielding heterogeneous structures. Residue preferences at the unnatural i + 1st, i + 3rd, and i + 4th positions of the HBS helix were determined by systematic CD, FT-IR absorption, 1D-, 2D-NMR spectral and computational analyses of Gly/Ala mutants at these positions, for their predictable application. Unlike in natural α-helices, Gly can be accommodated in these HBS helices and, when placed at the i + 1st position, yield the highest known helicities for single α-helical-turn models. Ala_i+1 rigidifies the helical turn and yields slightly reduced α-helicities than Gly_i+1. Ala_i+3 (non-Gly_i+3) is necessary, as Gly_i+3 yields a mixture of turns. Gly_i+4 is essential for probability of propagation downstream from the constrained α-helix. Ala_i+4 precludes such propagation and twists the i + 3–i + 4 tertiary amide bond away from the helix axis. The θ_n→π*MRE value (∼−25.3 × 10³ deg cm² dmol^–1 at pH 7) for 1a (Moc[GFA]G-OMe), the analogue containing the most preferred residues in all positions, is presented as a reference for ideal single-turn α-helices with the HBS constraint. The primary reason for the difference between this value and that of natural infinite α-helices is found to be the unavoidable greater interpeptide distance, due to the presence of the large HBS motif in the former. Current results facilitate chemical and biological investigations by providing crucial design criteria for predictable access to HBS-constrained stable single-turn α-helices. By setting a paradigm for an ideal α-helical turn with the highest helicity, our findings propose a strategy for increasing the helicity of mimetics of biologically important single α-helical turns and thus optimizing their relevant functionality.

Experimental Section

General Details

The protected amino acids were purchased from G.L. Biochem Ltd., China, and were used without additional purification. All reactions were carried out in oven-dried round-bottom flasks. Flash chromatography was performed on silica gel standard grade (100–200 mesh) for purification of all of the intermediate compounds. Nuclear magnetic resonance (NMR) spectra of compounds were recorded on a Bruker-AV400 spectrometer (Bruker Co., Faellanden, Switzerland). ¹H NMR spectra were recorded in CDCl₃ or 40% CDCl₃/CD₃CN or 10% D₂O/H₂O on a Bruker 400 MHz (100 MHz for ¹³C) spectrometer. Chemical shifts were reported relative to tetramethylsilane (TMS) (δ 0.00 ppm) for ¹H NMR and ¹³C NMR, respectively. High-resolution mass spectra (HRMS) were recorded on a Xevo TQ-XC mass spectrometer (Water, Massachusetts). Multiplicities are indicated using the following abbreviations: s (singlet), d (doublet), dd (doublet of doublets), dt (doublet of triplet), t (triplet), q (quartet), quin (quintet), sext (sextet), hept (heptet), m (multiplet), and bs (broad singlet).

CD Spectroscopic Analyses

CD spectra were recorded in the far-UV range, from 185 to 270 nm, at 295 K with a scan speed of 50 nm min^–1 and with each reading averaged over three scans. Spectral baselines were obtained under conditions analogous to those for the samples. The blank solvent spectra for each solution were recorded under the same conditions. Solutions were prepared by weighing out the peptide in a volumetric flask and adding the solvent for dilution up to the marks, ensuring the dissolving of the peptide, followed by filtering of the solution through a 0.2 μm polyvinylidene difluoride (PVDF) membrane filter (Pall India Pvt. Ltd., Mumbai). All of the spectra were baseline-corrected, and θ values were recorded in mdeg units. Each data point was then converted to uniform-scale molar ellipticity θ (deg cm² dmol^–1) values. The corresponding mean molar residue ellipticity (θ_MRE) values were then calculated by the equation θ/n (where n is the number of peptide bonds in the peptide), and the θ_MRE (deg cm² dmol^–1) values were plotted as a function of corresponding λ_nm. Temperature-dependent CD experiments were performed by varying the temperature of the samples using the JASCO Peltier instrument, allowing 15 min of equilibration time at each temperature before recording each data point.

FT-IR Spectroscopic Analyses

IR spectra were recorded using a Bruker spectrophotometer. A cell with path length 0.1 cm (with KCl window) was used for the solution measurements. All of the spectra were baseline-corrected with respect to the blank solvent (spectrograde CH₃CN purchased from Merck) with a minimum of 16 scans being signal-averaged. For acquiring the FT-IR for the deuterated analogues of 1a and 1b, the peptides were dissolved (5 mM concentration) in 100% CD₃OD (99.8% deuterated) and kept standing for 1 h and ¹H NMR were taken to ensure the complete exchange of NHs to NDs. Then, CD₃OD was removed under reduced pressure until complete dryness. Further, the peptides were redissolved in CH₃CN (2 mM concentration) and the spectra were recorded.

Procedure for Energy Minimization

Energy minimization was carried out on all three 13₇3 analogues 1a–c, starting from the structure that is built based on the 2D ROESY constraints. The interproton distances are taken care of based on the ROE intensities. The molecule was then uploaded into the ATB topology builder to generate the initial optimized geometry using the semiempirical QM theory (SCF level of theory) and MOPAC charges. Bonded and van der Waals parameters were taken from the GROMOS 54A7 parameter set. All three molecules were then placed in a suitable cubic box with the box edge adjusted to 1 Å from the peptide’s periphery. The box was filled with a suitable number of pre-equilibrated SPC water molecules. Then, energy minimization was done using the steep (steepest descent minimization) algorithm for all three molecules, which were restrained to their initial coordinates with a force constant of 100 kJ mol^–1.

General Procedure for the Cyclization Reaction

(Scheme S24, SI) MeOH (3.0 mL) was added to a mixture of 10a (300 mg, 0.43 mmol) and Pd–C (0.1 mol %) in a sealed round-bottom flask kept under a H₂ atmosphere and stirred for 20 min by which time TLC indicated complete consumption of the starting material. Filtering the mixture through Whatman-40 filter paper and concentration of the organic filtrate gave the N- and C-terminal double deprotected derivative of 10a as a viscous liquid 11a (208.8 mg, 0.43 mmol). Next, 11a was dissolved in dry CH₃CN (1.0 mM) (420.0 mL) followed by addition of 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC) (410.6 mg, 2.15 mmol), hydroxybenzotriazole (HOBT) (174.1 mg, 1.29 mmol), and N,N-diisopropylethylamine (DIPEA) (375.0 μL, 2.15 mmol) under a N₂ atmosphere. The mixture was stirred for a further 7 h. Removal of the solvent resulted in a residue that was dissolved in dichloromethane (DCM) (30.0 mL) and washed with water (2 × 15.0 mL), 1 N HCl (2 × 15.0 mL), and saturated NaHCO₃ (2 × 15.0 mL). The organic layer was dried over anhyd. Na₂SO₄ and concentrated to get a residue, which was subjected to purification by silica gel flash column chromatography.

Methyl (5S,8S)-5-Benzyl-10-(2-methoxy-2-oxoethyl)-8-methyl-3,6,9-trioxo-1,4,7,10-tetraazacyclotridecane-1-carboxylatewhite (1a)

White solid (89.35 mg, 0.19 mmol, 46% yield); (TLC—EtOAc/methanol (9:1)—R_f = 0.7); ¹H NMR (400 MHz, 40% CDCl₃ in CD₃CN, 5 mM) δ ppm: major conformation: 7.33–7.21 (m, 5H, H^AroPhe), 6.61 (d, J = 9.29 Hz, 2H, H^NPhe + Ala), 4.75 (dt, J = 16.57, 10.07 Hz, 1H, H^αAla₃), 4.50 (d, J = 15.11 Hz, 1H, H^αGly₁), 4.44 (d, J = 19.05 Hz, 1H, H^αGly₄), 4.31 (q, J = 6.12 Hz, 1H, H^αPhe₂), 4.11 (t, J = 13.52 Hz, 1H, Hprp^c), 3.77 (d, J = 19.12 Hz, 1H, H^αGly₄), 3.71 (s, 3H, H^MeOMe), 3.66 (d, J = 9.27 Hz, 1H, H^MeMoc), 3.54 (d, J = 16.17 Hz, 1H, H^αGly₁), 3.45–3.32 (m, 1H, Hprp^a), 3.30–3.16 (m, 1H, Hprp^a), 3.04–2.96 (m, 2H, H^βPhe₂), 2.52 (d, J = 14.53 Hz, 1H, Hprp^c), 2.29–2.17 (m, 1H, Hprp^b), 1.87–1.74 (m, 1H, Hprp^b), 1.33 (d, J = 6.11 Hz, 1H, H^βAla₃), 1.21 (d, J = 6.37 Hz, 3H, H^βAla₃); minor conformation: 7.17–7.09 (m, 0.11H, H^NPhe₂), 7.07–6.98 (m, 0.19H, H^NPhe₂), 6.91–6.83 (m, 0.11H, H^NAla₃), 6.76–6.67 (m, 0.19H, H^NAla₃); ¹³C NMR (100 MHz, 40% CDCl₃ in CD₃CN, 5 mM) δ ppm: (mixtures of major and minor conformers) 173.2, 171.5, 170.9, 170.1, 136.7, 129.7, 129.1, 127.6, 58.1 (C^αPhe₂), 52.9 (C^MeMoc), 52.6 (C^MeOMe), 52.5 (C^αGly₁), 48.2 (C^αGly₂), 46.0 (Cprp^a), 44.4 (C^αAla₃), 42.9 (Cprp^c), 36.8 (C^βPhe₂), 16.9 (C^βAla₃); FT-IR (Ṽ cm^–1) (CH₃CN, 2 mM): 2978, 2868, 1749, 1700, 1657, 1528, 1472, 1358, 1315, 1290, 1220; HRMS m/z calcd for C₂₂H₃₀N₄O₇Na 485.2012, found 485.2014.

Methyl (5S,8S)-5-Benzyl-10-(2-methoxy-2-oxoethyl)-2,8-dimethyl-3,6,9-trioxo-1,4,7,10-tetraazacyclotridecane-1-carboxylate (1b)

White solid (77.6 mg, 0.16 mmol, 38% yield); (TLC—EtOAc/methanol (9:1)—R_f = 0.8); ¹H NMR (400 MHz, 40% CDCl₃ in CD₃CN, 5 mM) δ ppm: major conformation: 7.36–7.20 (m, 5H, H^AroPhe₂), 6.68 (d, J = 9.86 Hz, 1H, H^NAla₃), 6.51 (s, 1H, H^NPhe₂), 4.78 (dt, J = 17.02, 6.70 Hz, 1H, H^αAla₃), 4.69–4.62 (m, 1H, H^αAla₁), 4.58 (d, J = 19.13 Hz, 1H, H^αGly₄), 4.25–4.17 (m, 1H, H^αPhe₂), 4.1 (t, J = 13.63 Hz, 1H), 3.76 (d, J = 18.99 Hz, 1H, H^αGly₄), 3.71 (s, 3H, H^MeOMe), 3.29–3.19 (m, 1H, Hprp^a), 3.21–3.17 (m, 1H, Hprp^a), 3.07–2.97 (m, 2H, H^βPhe₂), 1.76–1.66 (m, 1H, Hprp^b), 1.28 (d, J = 7.07 Hz, 3H, H^βAla₁), 1.24 (d, J = 6.27 Hz, 3H, H^βAla₃); minor conformation: 7.17–7.12 (m, 0.36H, H^NPhe₂), 7.12–7.07 (m, 0.36H, H^NAla₃); ¹³C NMR (100 MHz, 40% CDCl₃ in CD₃CN, 5 mM) δ ppm: (mixtures of major and minor conformers) 174.3, 172.0, 171.8, 171.7, 170.9, 138.3, 137.6, 130.6, 130.5, 130.3, 130.9, 129.8, 129.7, 128.4, 128.2, 59.2 (C^αPhe₂), 58.3 (C^αAla₁), 53.5 (C^MeMoc), 53.4 (C^MeOMe), 48.9 (C^αGly₄), 45.0 (C^αAla₃), 43.5 (Cprp^c), 41.4 (Cprp^a), 37.6 (C^βPhe₂), 28.9 (Cprp^b), 17.4 (C^βAla₁), 15.9 (C^βAla₃); FT-IR (Ṽ cm^–1) (CH₃CN, 2 mM): 2992, 2946, 1750, 1691, 1654, 1527, 1507, 1472, 1440, 1402, 1375, 1302, 1210, 1112, 1059, 1003; HRMS m/z calcd for C₂₃H₃₂N₄O₇Na 499.2169, found 499.2166.

Methyl (5S,8S)-5-Benzyl-10-(1-methoxy-1-oxopropan-2-yl)-8-methyl-3,6,9-trioxo-1,4,7,10-tetraazacyclotridecane-1-carboxylate (1c)

White solid (45 mg, 0.09 mmol, 22% yield); (TLC—EtOAc/methanol (9:1)—R_f = 0.8); ¹H NMR (400 MHz, 40% CDCl₃ in CD₃CN, 5 mM) δ ppm: major conformation: 7.14 (d, J = 8.49 Hz, 1H, H^NPhe₂), 7.35–7.23 (m, 5H, H^AroPhe), 7.16 (d, J = 9.49 Hz, 1H, H^NAla₃), 4.76 (q, J = 6.72 Hz, 1H, H^αAla₃), 4.65 (q, J = 7.90 Hz, 1H, H^αPhe₂), 3.99 (d, J = 13.44 Hz, 1H, H^αGly₁), 3.86 (q, J = 6.88 Hz, H^αAla₄), 3.74 (s, 3H, H^MeOMe), 3.70–3.67 (m, 1H, H^αGly₁), 3.62 (s, 3H, H^MeMoc), 3.38–3.28 (m, 1H, Hprp^c), 3.27–3.18 (m, 1H, Hprp^a), 3.13–3.04 (m, 1H, H^βPhe₂), 3.03–2.97 (m, 1H, Hprp^c), 2.97–2.91 (m, 1H, H^βPhe₂), 1.91–1.79 (m, 1H, Hprp^b), 1.72–1.60 (m, 1H, Hprp^b), 1.40 (d, J = 6.58 Hz, 3H, H^βAla₄), 0.94 (d, J = 6.49 Hz, 3H, H^βAla₃); minor conformation: 6.99–6.91 (m, H^NAla₃), 6.84–6.74 (m, H^NAla₃); ¹³C NMR (100 MHz, 40% CDCl₃ in CD₃CN, 5 mM) δ ppm: (mixtures of major and minor conformers) 173.3, 172.0, 171.4, 171.3, 158.6, 138.3, 130.54, 130.48, 129.9, 129.7, 128.2, 128.0, 57.6 (C^αAla₄), 57.4, 56.0 (C^αPhe₂), 54.3 (C^αGly₁ + C^MeOMe), 53.0 (C^MeMoc), 48.1 (Cprp^c), 47.5 (Cprp^a), 46.9 (C^αAla₃), 38.6 (C^βPhe₂), 29.5 (Cprp^b), 18.0 (C^βAla₃), 16.0 (C^βAla₄); FT-IR (Ṽ cm^–1) (CH₃CN, 2 mM): 3092, 3037, 2603, 2265, 2117, 1880, 1741, 1701, 1639, 1534, 1477, 1245, 1193; HRMS m/z calcd for C₂₃H₃₂N₄O₇Na 499.2169, found 499.2168.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsomega.0c01277.

• Detailed synthetic procedures including scales, equivalents of reagents, schemes, and yields; detailed procedures for acquisition of NMR (1D and 2D), CD, FT-IR spectra, and for energy minimization; tables and figures of all of the CD (concentration, temperature, solvent-dependent, etc.); 1D and 2D-NMR spectra, ROE figures and tables; relevant figures, tables, and diagrams to determine the 3D structure of the HBS-constrained macrocyclic peptides (PDF)

The authors declare no competing financial interest.