Reagents and solvents were purchased from commercial sources and used without further purification. Column chromatography was carried out on silica gel (300~400 mesh). ¹H NMR spectra were recorded at 400 MHz and 600 MHz on a Bruker-400 spectrometer and JEOL-400 and 600 spectrometers at ambient temperature. ¹³C NMR spectra were measured at 100 MHz and 150 MHz on the same spectrometers. Chemical shifts are reported in parts per million downfield from TMS (tetramethylsilane). Coupling constants in ¹H NMR are expressed in Hertz. Electrospray ionization high resolution mass spectra (ESI-HRMS) were acquired using a waters LCT Premier XE spectrometer (Waters, Milford, MA, USA).

The models for 2a and 2b were optimized with the revPBE-D3 (Zhang and Yang, 1998; Grimme et al., 2010) functional and dispersion correction using the Amsterdam Density Functional (ADF) (Fonseca Guerra et al., 1998; te Velde et al., 2001)¹ software package. The triple-zeta with polarization functions (TZP) basis set was used while keeping the core 1s electrons fixed in the oxygen, nitrogen, and carbon atoms (van Lenthe and Baerends, 2003). The revPBE-D3 functional and dispersion correction were used due to its previous treatment of similar tetrapeptides (Zhang et al., 2019).

The synthesis of tetrapetides 1 and 2 has been reported by us (Zhang et al., 2019; Tang et al., 2020). Peptides 1′ and 2′ were prepared based on standard amide/peptide coupling.

Figure 1 shows the general steps and conditions for synthesizing hexapeptides 1a, 2a, and 2b. Coupling I and II, which were prepared by coupling the methyl ester of D- or L-β-homo-phenyalanine with 2-isopentyloxy-5-nitrobenzoic acid, and Boc-protected glycine with the methyl ester of 5-amino-2-isopentyloxybenzoic acid, results in oligomer III. Hydrolyzing the methyl ester gives IV, which is coupled with the L-leucine derived amide to give V. Subjecting V to catalytic hydrogenation results in VI, followed by coupling with acetyl-L-valine to give peptides 1a, 2a, and 2b. The detailed synthetic steps for preparing the intermediates and final products, along with the corresponding analytical data are included in the Supplementary Materials.

Results from our recent study indicate that hybrid tetrapeptides 1 and 2, along with eight other homologous hybrid peptides (Tang et al., 2020), persistently fold into a doubly H-bonded hairpin conformation containing an expanded β-turn that is very resilient toward incorporating different α- and β-amino acid residues into the central β/α dipeptide loop. Such a turn motif is capable of accommodating a variety of β/α dipeptide sequences that otherwise could not be introduced into a β-turn. For example (Figure 2A), glycine and β-alanine, i.e., homoglycine, two conformationally most flexible α- and β-amino acid residues, are found in tetrapeptide 1 which adopts a well-defined hairpin conformation. Replacing the β-alanine residue of 1 with other β-amino acid residues having side chains results in hybrid tetrapeptides such as 2 that adopts a hairpin conformation with enhanced stability.

Each of tetrapeptides 1 and 2, like other hybrid tetrapeptides of this series, has two aromatic γ-amino acid (γAr) residues flanking the central β/α dipeptide segment. The vital role of the γAr residues in driving the folding of 1 and 2 is further demonstrated by examining the folding of tetrapeptides 1′ and 2′ (Figure 2B), which share the same β/α dipeptide segments with 1 and 2, respectively, but the latter two have two α-amino acid residues, i.e., L-Val and L-Leu, that flank the β/α dipeptide segment. If folded, tetrapeptides 1′ and 2′ could also adopt doubly H-bonded hairpin conformations as shown in Figure 2B.

The ¹H NMR spectra of 1 and 1′, and 2 and 2′ recorded at 25 mM were compared to those recorded at 1 mM in CDCl₃. Table 1 shows the difference in the chemical shifts (Δδ_NH) of the amide protons of the four peptides at the two concentrations. In the folded conformations of 1 and 2, protons a and d, like b and e, are intramolecularly H-bonded and exhibit either small upfield shifts or insignificant downfield shifts at high vs. low concentrations; while the signals of protons c, which are not intramolecularly H-bonded, shift noticeably downfield with increasing concentration (Tang et al., 2020). In contrast, all of the amide proton resonances of peptides 1′ and 2′ recorded at 25 mM show downfield shifts relative to those at 1 mM, with the signals of protons a, c, and d showing significant shifts, while those of protons b undergoing small shifts. These observations suggest that among the amide protons of 1′ and 2′, only protons b are intramolecularly H-bonded. The fact that protons a and d of 1′ and 2′ are not intramolecularly H-bonded indicates that 1′ and 2′ do not fold into the hairpin conformation as shown in Figure 2B.

The conformations of tetrapeptides 1′ and 2′ were examined with 2D (NOESY) spectroscopy (Supplementary Figures S4, S5). Except for NOEs between protons b and c, and c and d, the spectrum of 1′ reveals no NOEs between protons a and d, i and l, or j and k, which would exist if a hairpin conformation existed. The spectrum of 2′ contains an NOE between protons b and c, with no NOEs between protons a and d, i and l, or j and k being observed. The fact that only proton b of 1′ or 2′ is intramolecularly H-bonded, along with the absence of NOEs between other remote protons, suggests that 1′ and 2′, being derived from replacing the aromatic γ-amino acid residues 1 and 2 with L-Val and L-Leu residues, cannot fold into hairpin conformations.

The above observations indicate that the doubly H-bonded γAr residues are indispensable in driving the folding of 1 and 2 into hairpin conformations. Without the γAr residues, peptides 1′ and 2′, although capable of forming intramolecular H-bonds involving protons a and d, fail to adopt hairpin conformations. The critical role played by the γAr residues on stabilizing these novel hairpins relies on the effective H-bonding capabilities offered by these structural units (Gong, 2007), which provides the energetic driving force for the observed persistent folding of these hybrid peptides.

Comparing tetrapeptides 1′ with 1, and 2′ with 2 revealed the critical importance of the doubly H-bonded γAr residues in ensuring the adoption of hairpin conformations by 1 and 2. Attaching additional amino acid residues to 1 or 2 results in a longer peptide that may fold into a hairpin with an enhanced stability due to the energetic contribution of added H-bond(s). Hexapeptide 1a (Figure 3), which is resulted from adding a pair of amino acid residues, L-Val and L-Leu, to the N and C termini of 1, respectively, were examined and compared to tetrapeptide 1.

The ¹H NMR spectrum of 1a recorded in CDCl₃ at 25 °C contains well-dispersed signals (Supplementary Figure 1), indicating that 1a, like 1, exists as a single discrete species with a defined conformation. At 1 mM in CDCl₃, the signals of protons a and d of 1a appear at 10.06 and 9.53 ppm, respectively, while the same protons of 1 are found at 9.65 and 9.39 ppm. The downfield shifts of protons a and d of 1a relative to those of 1 at the same concentration indicate that the H-bonds involving protons a and d of the former are stronger than those of the latter, which suggests that, hexapeptide 1a, with one additional H-bond contributed by the L-Val/L-Leu pair, may very likely fold into a triply H-bonded conformation that is more stable than the doubly H-bonded hairpin of 1.

Comparing the difference in the chemical shifts of amide protons at 25 mM and 1 mM in CDCl₃ reveals that the amide proton resonances of 1a follow the same trend as shown by those of 1 (Supplementary Table 1). The signal of proton c of 1 or 1a undergoes the largest downfield shift (~0.7 ppm) upon increasing the concentration from 1 to 25 mM, suggesting that proton c of 1a, like that of 1 (Tang et al., 2020), is intermolecularly H-bonded. In contrast, the signals of amide protons a, b, d, and e of 1a exhibit very small (<0.01 ppm) upfield shifts, indicating that protons a and d, like b and e, are intramolecularly H-bonded. These observations suggest that hexapeptide 1a, like tetrapeprtide 1, folds into a hairpin conformation that is stabilized by H-bonds involving protons a and d, and further reinforced by a H-bond involving proton f. Indeed, proton f undergoes an upfield shift of 0.274 ppm from 1 to 25 mM, suggesting that it is intramolecularly H-bonded.

The H-bonding interactions involving protons a and d were further examined by monitoring the chemical shifts of amide protons a and d of peptides 1 and 1a in CDCl₃ containing DMSO-d₆. Interestingly, the signals of protons a and d shift differently with increasing ratios of DMSO (Figure 4). The resonances of protons a of both 1 and 1a first shift upfield and then move downfield with as the ratio of DMSO increases (Figure 4A). In contrast, the signals of protons d of both 1 and 1a show overall linear downfield shifts with increasing DMSO ratio (Figure 4B).

The different shifts of amide protons a and d toward increasing solvent polarity can be explained by the folded conformations of 1 and 1a (Figure 3). Proton d is involved in an “intraturn”, N–H(i) → O=C(i −3) hydrogen bond that is part of the 11-atom, intramolecularly H-bonded ring that constitutes the expanded β-turn in the hairpin conformation of 1 or 1a. Such a H-bond, with a H∙∙∙O distance of over 2.0 Å, is slightly longer and thus weaker than typical H-bonds. As a result, proton d of 1 or 1d is more accessible to solvent molecules than proton a. With increasing proportion of DMSO, proton d becomes increasingly H-bonded with DMSO molecules, which results in the destabilization of the folded (hairpin) conformation. Such an overall destabilization of the hairpin conformation in turn weakens the H-bond involving proton a, which is reflected by the initial upfield shift of the signal of proton a. As the H-bond becomes further weakened, proton a also becomes more exposed to solvent molecules and engages in increasing H-bonding interaction with DMSO molecules, which leads to the downfield shift of its signal.

Comparing the shifts of amide protons a of 1 and 1a reveals another interesting trend. As show in Figure 4A, the upfield shift shown by the signal of proton a of 1 is reversed at ~5% DMSO, while that of 1a is reversed at 15% DMSO, indicating that proton a of 1a is more resistant toward increasing solvent polarity than that of 1, i.e., the H-bond involving proton a in 1a is stronger than that in 1. The stronger H-bond involving proton a of 1a is mostly like due to the higher overall stability of the hairpin conformation of 1a than that of 1. The enhanced stability shown by the folded structure of 1a can be explained by the energetic contribution of H-bonding involving proton f and the terminal amide C = O group.

The folded conformation of hybrid peptide 1a is confirmed by 2D NOESY spectra. As shown in Figure 5, the NOEs observed with 1a include those between protons a and m, and h and n, which indicate the H-bonded alignment of the two aromatic γ-amino acid residues and the L-Val and L-Leu residues. In addition, NOEs between protons c and j, c and k, c and d, and d and l demonstrate the presence of a well-defined loop. These multiple NOEs confirm that hybrid peptide 1a folds into a hairpin conformation similar to those observed with hybrid tetrapeptides 1 and 2 (Tang et al., 2020).

Results from our recent study demonstrate that replacing the β-alanine residue of 1 with other β-amino acid residues such as β-homoPhe gives hybrid tetrapeptides including 2 that fold into expanded β-turns with enhanced stabilities (Tang et al., 2020). It is expected that the β-turn of 2, like that of 1, should also accommodate additional amino acid residues, resulting in longer β-hairpins. Attaching L-Val and L-Leu residues to the N and C termini of 2 results in hexapeptide 2a. Replacing the L-β-homoPhe residue of 2a with D-β-homoPhe gives hexapeptide 2b (Figure 6). By comparing the folding of 2, 2a, and 2b, and the stabilities of the folded structures, we intend to probe the compatibility of the chirality of the β-homoPhe residue in the dipeptide segment of the expanded β-turn with that of the two terminal L-α-amino acid residues.

Table 2 lists the difference in the chemical shifts of amide protons a, b, c, d, and e of peptides 2, 2a, and 2b, along with the chemical shifts of amide protons f and g of 2a and 2b, measured at 25 mM and 1 mM. The amide protons of 2 and 2b show the same overall change in their chemical shifts measured at high and low concentrations, suggesting that 2b, like 2, also folds into a well-defined hairpin conformation. In contrast, the amide proton resonances of 2a show noticeable difference in that proton c does not undergo as large a downfield shift as those shown by protons c of 2 and 2b. In addition, proton d of 2a shows a significant downfield shift from 1 to 25 mM, which contrasts the negligible shifts observed with protons d of 2 and 2b. These observations imply that, compared to those of 2 and 2b, proton d of 2a is more exposed and is available for intermolecular H-bonding, which in turn suggests that the expanded β-turn of 2a involving H-bonded proton d might be partially twisted. Such a partial twisting or deformation may be resulted from the incompatibility of the L-homoPhe with the terminal L-Val and L-Leu residues of 2a.

To assess the relative stabilities of the folded conformations of 2, 2a, and 2b, the chemical shifts of protons a and d of these three hybrid peptides in CDCl₃ containing DMSO-d₆ were compared. Similar to what is observed with 1 and 1a (Figure 4), the signals of protons a of 2, 2a and 2b also first shift upfield with increasing proportion of DMSO, followed by shifting downfield as the ratio of DMSO further increases (Figure 7A). The resonances of protons a of the three peptides, however, undergo transitions from upfield to downfield shifts at different percent of DMSO, with peptide 2 showing its transition at ~6% DMSO, 2a at ~8% DMSO, and 2b at ~25% DMSO. Thus, proton a of 2b is the least responsive toward increasing solvent polarity, which indicates that the folded conformation of 2b is the most stable among those of the three peptides. The high stability of folded 2b is also demonstrated by the upfied and then down field shift of proton d of this peptide (Figure 7B), which contrasts the consistent downfield shifts observed with the resonances of protons d of the other hybrid peptides, indicating that the H-bond involving proton d is greatly enhanced in the folded conformation of 2b.

In contrast, the stability of folded hexapeptide 2a is similar to that of tetrapeptide 2 and is much less stable than those of hexapeptides 2b and 1a (Figures 4A, 7A). The strong H-bonding and high stability of 2b, and the much lower stability of 2a, suggest that D-homoPhe residue in the H-bonded loop of 2b is more compatible with the L-Val and L-Leu residues, while L-homoPhe of 2a is much less compatible. Therefore, a hetero-chiral combination of the β-amino acid residue in the expanded β-turn and the α-amino acid residues in the β-strands seems to favor the nucleation and stabilization of β-hairpins. For example, an expanded β-strand with a D-β-amino acid residue should promote oligopeptides of L-α-amino acids to pair into a β-sheet.

NOESY spectra of 2a and 2b provide additional insights into the folding of these two hexapeptides. As shown in Figure 8A, the NOEs observed with 2a include those between protons d and i, and h and n, which indicate the alignment of the two γAr residues and the L-Val and L-Leu residues. In addition, NOEs between protons c and j, c and k, c and d, and d and l demonstrate the presence of the H-bonded loop. The observed NOEs suggest that hybrid peptide 2a adopts an overall hairpin conformation that includes a H-bonded loop along with the expected alignment of γAr and α-amino acid residues. In comparison to those of 2a, NOEs with significantly stronger intensities are revealed by the NOESY spectrum of 2b (Figure 8B). Strong NOEs that indicate the H-bonded alignment of the two γAr residues and the L-Val/L-Leu residues are clearly observed between protons a and m, a and n, d and i, and h and n. NOEs between protons c and j, c and k, c and d, d and j, and d and l are consistent with the presence of the 11-atom H-bonded ring that constitutes the expanded β-turn. The different numbers and strengths of the NOEs detected for 2a and 2b are consistent with the above conclusion on the different stabilities of the two folded structures. The numerous strong NOEs observed with 2b is consistent with a compact, tightly folded conformation.

Finally, hexapeptides 2a and 2b were computationally optimized and compared. The optimized structures are shown in Figure 9. Modeling of 2a with revPBE-D3 revealed N-H∙∙∙O hydrogen bond distances for protons a, d, and f to be 1.79, 2.08, and 1.91 Å, respectively. Measurements of the N-H∙∙∙O hydrogen bond distances in 2b show that H-bonds involving protons a, d, and f shorten to 1.78, 1.98, and 1.88 Å, respectively, which suggest strengthened H-bonds. The strengthening of the hydrogen bonds in 2b relative to 2a, assessed through the N-H∙∙∙O hydrogen bond distances, coupled with 2a having a larger buckling within the turnabout of the L-homoPhe residue relative to 2b supports the experimental findings that 2b holds a greater overall stability than 2a.