Usual solvents were purchased from commercial sources, dried, and distilled by standard procedures. Pure compounds were obtained after liquid chromatography using Merck silica gel 60 (40–63 μm). TLC analyses were performed on silica gel 60 F250 (0.26 mm thickness) plates. The plates were visualized with UV light (λ = 254 nm) or revealed with a 4% solution of phosphomolybdic acid or ninhydrin in EtOH.

NMR spectra were recorded on a Bruker AMX 200 (¹H, 200 MHz; ¹⁹F, 188 MHz), an ultrafield Bruker AVANCE 300 (¹H, 300 MHz, ¹³C, 75 MHz), a Bruker AVANCE 400 (¹H, 400 MHz, ¹³C, 100 MHz, ¹⁹F, 376 MHz), a Bruker AVANCE I 600 MHz (¹H, 600 MHz, ¹³C, 150 MHz, ¹⁹F, 564 MHz) equipped with a z-gradient TCI and ¹⁹F-QCI, or a Bruker AVANCE III 800 MHz (¹H, 800 MHz, ¹³C, 200 MHz) spectrometers, equipped with a z-gradient TCI cryoprobe.

Chemical shifts (δ) are in ppm and downfield from Me₄Si (δ = 0.0 ppm) with the solvent resonance as the internal standard (¹H NMR, CDCl₃: δ = 7.26 ppm, CD₃OD and CD₃OH: δ = 3.31 ppm; ¹³C NMR, CDCl3: δ = 77.16 ppm, CD₃OD and CD₃OH: δ = 49.00 ppm). the following abbreviations are used: singlet (s), doublet (d), doublet of doublet (dd), triplet (t), quadruplet (q), multiplet (m), and broad singlet (bs).

Melting points were determined on a Kofler melting point apparatus. High-resolution mass spectra (HRMS) were obtained using a TOF LCT Premier apparatus (Waters), with an electrospray ionization source. The purity of compounds was determined by HPLC using a WATERS gradient system (pump + controller E 600, UV detector PDA 2996, autosampler 717) on a XSelect column (C18, 2.1 mm × 150 mm–3.5 μm), mobile phase, MeCN/H₂O + 0.1% formic acid (gradient 5–100% in 20 min), detection at 257 nm. Preparative HPLC were performed on Agilent Infinity II.

Compounds 3–4, 8–9, and 14 were subjected to conformational studies by NMR in CD₃OH. When signals were sufficiently resolved and disperse, vicinal ³J_HN–Hα coupling constants, H^α-HN ROE correlations, temperature coefficient (Δδ_HN/ΔT) of the amide protons and ¹H_α and ¹³C_α chemical shift deviations (CSD), were examined to analyze backbone conformational propensities.

All the protocols of synthesis and characterization data are given in Supplementary Material.

9a and 9b three-dimensional structures were generated using the program MarvinSketch 6.2.1 from ChemAxon¹. In both initial conformers, the torsion angle between the piperidine and proline (Cα-N-C3-C2) has a value of −60°, so that the proline Hα points toward piperidine H2, as observed in NMR experiments. The torsion angle around the amide bond between Leu-3 and Pip-4 (Cα-C-N-C2) was manually adjusted to −170° and −5° in 9a and 9b, respectively. Each conformer was placed in a cubic simulation box so that the minimum distance between the solute and the cube sides was 1.4 nm. Afterward, both simulation boxes were filled with explicit methanol molecules. Molecular dynamics simulations were performed by using the GROMACS 2019.1 package (Abraham et al., 2015) with the Generalized AMBER Force Field (GAFF; Wang et al., 2004) for both solute and methanol solvent. The length of the covalent bonds involving hydrogens was kept constant using the LINCS procedure (Hess, 2008). Lennard-Jones potentials were cut-off at 1.2 nm and electrostatic interactions were treated using the smooth PME method (Essmann et al., 1995). After two short simulations of 1 ns each to bring the system temperature and pressure around T = 300 K and P = 1 bar, each system was simulated during 200 ns in the isothermal-isobaric ensemble using the Nose-Hoover (Nosé, 1984; Hoover, 1985) and Parrinello-Rahman (Parrinello and Rahman, 1981) coupling methods. Molecular coordinates were saved every 20 ps for subsequent analyses by GROMACS tools, including gmx angle for extracting torsion angle values.

The LUVs were composed of a mixture of DOPC/DOPS in a 7:3 molar ratio. Stock solutions of DOPC and DOPS in chloroform at concentrations of 20–30 mM were mixed in a glass tube. The solvent was evaporated with dry nitrogen gas yielding a lipid film that was subsequently kept in a vacuum desiccator for 1 h. Lipid films were hydrated during at least 30 min in the 10 mM Tris, 100 mM NaCl buffer at pH 7.4. The lipid suspensions were subjected to 10 freeze-thaw cycles, at temperatures of approximately −190 and 50°C, respectively, and subsequently extruded 19 times through a mini-extruder (Avanti Alabaster, AL, United States) equipped with polycarbonate membranes (at 200 nm cut-off). Calcein-containing vesicles were obtained by adding 70 mM calcein to the hydration buffer and free calcein were removed from the LUVs using size exclusion chromatography (Sephadex G50-fine), using the hydration buffer (10 mM Tris, 100 mM NaCl, pH 7.4) to as the mobile phase. The phospholipid content of lipid stock solutions and vesicle preparations was determined by assessing inorganic phosphate according to Rouser (Rouser et al., 1970).

Human islet amyloid polypeptide, purchased from Bachem, was dissolved in pure hexafluoro-isopropanol (HFIP) at a concentration of 1 mM and incubated for 1 h at room temperature to dissolve any preformed aggregates. Next, HFIP was evaporated with dry nitrogen gas followed by vacuum desiccation for at least 3 h. The resulting peptide film was then dissolved in DMSO to obtain stock solutions of hIAPP (0.2 mM) and stock solutions of compounds to test were dissolved in DMSO (10, 1, and 0.1 mM). The concentration of DMSO was kept constant at 3% (v/v) in the final volume of 200 μL. Thioflavin-T binding assays were used to measure the formation of fibrils in solution or in the presence of membranes over time. A plate reader (Fluostar Optima, BMG Labtech) and standard 96-wells flat-bottom black microtiter plates in combination with a 440 nm excitation and 480 nm emission filters were used. The ThT assay was started by adding 5 μL of a 0.2 mM hIAPP stock solution to a mixture of 10 μM ThT (obtained from Sigma) and 10 mM Tris/HCl, 100 mM NaCl at pH 7.4 containing 1 μL of stock solutions of compound to test. For the experiments in the presence of membranes, the fluorescence assays were started by adding 5 μL of a 0.2 mM hIAPP (5 μM peptide) to 195 μL of a mixture of 10 μM ThT, DOPC/DOPS (7:3) LUVs, with a peptide:lipid ratio of 1:20. The concentration of IAPP was held constant at 5 μM for all experiments and inhibitors were added to yield compound/IAPP ratios of 10/1, 1/1, and 0.1/1 (only at ratio 1/1 for the assays in membrane). The ThT assays were performed in triplicate and between 2 and 4 times on different days, with the same batch of peptide. The ability of compounds to inhibit IAPP aggregation was assessed considering the time of the half-aggregation (t1/2) and the intensity of the experimental fluorescence plateau (F), both values were obtained by fitting the obtained kinetic data to a Boltzmann sigmoidal curve using GraphPad Prism 5. The relative extension/reduction of t1/2 is defined as the experimental t1/2 in the presence of the tested compound relative to the one obtained without the compound and is evaluated as the following percentage: [t1/2 (hIAPP + compound) − t1/2 (hIAPP)]/t1/2 (hIAPP) × 100. The relative extension/reduction of the experimental plateau is defined as the intensity of experimental fluorescence plateau observed with the tested compound relative to the value obtained without the compound and is evaluated as the following percentage: (FhIAPP + compound – FhIAPP)/FhIAPP × 100. The curves of the tested compounds are fitted to a Boltzmann sigmoidal model, normalized to the control experiment.

Representative curves of ThT fluorescence assays over time showing hIAPP aggregation (5 μM) in the absence and in the presence of compounds 1–15 at compound/hIAPP ratios of 10/1, 1/1, and 0.1/1 are shown in Supplementary Material (Supplementary Figure 34). Representative curves of ThT fluorescence assays over time showing hIAPP aggregation (5 μM) in the presence of membranes, in the absence and in the presence of hairpins 8, 10, 13, and 14 at compound/hIAPP ratios of 1/1 are shown in Supplementary Material (Supplementary Figure 35).

Aβ_1–42 was purchased from Bachem and ThT was obtained from Sigma. The peptide was dissolved in an aqueous 1% ammonia solution to a concentration of 1 mM and then, just prior to use, was diluted to 0.2 mM with 10 mM Tris–HCl and 100 mM NaCl buffer (pH 7.4). Stock solutions of compounds to test were dissolved in DMSO with the final concentration kept constant at 0.5% (v/v). ThT fluorescence was measured to evaluate the development of Aβ1-42 fibrils over time using a fluorescence plate reader (Fluostar Optima, BMG Labtech) with standard 96-well black microtiter plates (final volume in the wells of 200 μL). Experiments were started by adding the peptide (final Aβ1-42 concentration equal to 10 μM) into a mixture containing 40 μM ThT in 10 mM Tris–HCl and 100 mM NaCl buffer (pH 7.4) with and without the compounds at different concentrations (100, 10, 1 μM) at room temperature. The ThT fluorescence intensity of each sample (performed in triplicate) was recorded with 440/480 nm excitation/emission filters set for 42 h performing a double orbital shaking of 10 s before the first cycle. The fluorescence assays were performed between 2 and 4 times on different days, with the same batch of peptide. The ability of compounds to inhibit Aβ1-42 aggregation was assessed considering the time of the half-aggregation (t1/2) and the intensity of the experimental fluorescence plateau (F), both values were obtained by fitting the obtained kinetic data to a Boltzmann sigmoidal curve using GraphPad Prism 5. The relative extension/reduction of t1/2 is defined as the experimental t1/2 in the presence of the tested compound relative to the one obtained without the compound and is evaluated as the following percentage: [t1/2 (Aβ + compound) − t1/2 (Aβ)]/t1/2 (Aβ) × 100. The relative extension/reduction of the experimental plateau is defined as the intensity of experimental fluorescence plateau observed with the tested compound relative to the value obtained without the compound and is evaluated as the following percentage: (FAβ + compound – FAβ)/FAβ × 100. Curves of the tested compounds are fitted to a Boltzmann sigmoidal model, normalized to the control experiment and represented in Supplementary Material.

Representative curves of ThT fluorescence assays over time showing Aβ_1–42 (10 μM) aggregation in the absence and in the presence of 1–3, 8, and 14 at compound/Aβ_1–42 ratios of 10/1, 1/1, and 0.1/1 are shown in Supplementary Material (Supplementary Figure 36).

Samples were prepared under the same conditions as in the ThT-fluorescence assay. Aliquots of hIAPP (5 μM in 10 mM Tris–HCl, 100 mM NaCl, pH 7.4 in the presence and absence of hairpins 3, 8, and 14 were adsorbed onto 300-mesh carbon grids for 2 min, washed and dried). The samples were negatively stained for 45 s on 2% uranyl acetate in water. After draining off the excess of staining solution and drying, the grids were observed using a JEOL 2100HC TEM operating at 200 kV with a LaB6 filament. Images were recorded in zero-loss mode with a Gif Tridiem energy-filtered-CCD camera equipped with a 2 k × 2 k pixel-sized chip (Gatan, Inc., Warrendale, PA, United States). Acquisition was accomplished with the Digital Micrograph software (versions 1.83.842, Gatan, Inc., Warrendale, PA, United States).

Leakage experiments were performed in standard 96-wells transparent microtiter plates using a plate reader (Spectrafluor, Tecan, Salzburg, Austria). Aliquots of 2.5 μL of molecules solution in DMSO at the desired concentration were added to 192.5 μL of 100 μM calcein-containing lipid vesicles in 10 mM Tris–HCl, 100 mM NaCl buffer at pH 7.4. The assay was then started by adding 5 μL of a 0.2 mM IAPP solution in DMSO or 2.5 μL DMSO only as control. Directly after addition of all components, the microtiter plate was shaken for 10 s. The plate was not shaken during the measurement. Fluorescence of calcein was measured from the top, every 5 min, using a 485 nm excitation filter and a 535 nm emission filter. The temperature during the measurement was 25 ± 3°C. The maximum leakage at the end of each measurement was determined by adding 2 μL of 10% Triton X-100 to a final concentration of 0.1% (v/v). The release of fluorescent dye was calculated according to the following equation:

L(t) is the fraction of dye released (normalized membrane leakage) at time t, Ft is the measured fluorescence intensity at time t, and F0 and F100 are the fluorescence intensities at times t = 0 and after addition of Triton X-100, respectively. All membrane leakage assays were performed three times, each in triplicate, on different days, using different IAPP stock solutions.

Sample preparation: commercial hIAPP was dissolved upon reception in pure HFIP at 1 mM and incubated for 1 h at room temperature, followed by an aliquoting and an immediate evaporation under nitrogen, then under vacuum and storage at −20°C. Just before performing real time kinetics, the dried peptide was reconstituted in 50 mM ammonium acetate buffer pH 3.7 at 100 μM. Compounds 3, 8, and 14 were dissolved in DMSO and added to the peptide solution to reach final concentrations 1 mM, 100 and 50 mM, corresponding, respectively, to ratios 10/1, 1/1, and 0.5/1 of compounds:hIAPP; the final concentration of DMSO in the ammonium acetate buffer being of 1% (v/v).

CE-UV experiments were carried out with a MDQ Instrument (SCIEX, Framingham, MA, United States) equipped with a UV detector. Detection was performed at 200 nm. Fused silica capillaries (50 μm id × 365 μm od) were purchased from Polymicro Technologies (Phoenix, AZ, United States). Silica capillaries were coated with a 0.2% polybrene solution. Briefly, solid polybrene was dissolved in water under heating at 50°C. The capillary was preconditioned with MeOH, NaOH 1 M, NaOH 0.1 M, and with H₂O. All flushings were done at 20 psi for 15 min. The capillary was then coated with the 0.2% polybrene solution for 15 min at 20 psi. The coated capillary was then flushed with the background electrolyte (BGE) at 20 psi for 45 min and equilibrated under −25 kV for 2 h. Polybrene-coated capillaries were 60 cm total length (49.8 cm to the detector). The BGE was a 50 mM ammonium acetate buffer, pH 3.7. The separation was carried out under −25 kV at 25°C. Samples were injected from the inlet by hydrodynamic injection at 0.8 psi for 10 s. After each run, the capillary was rinsed for 5 min with 1 M NaOH, 5 min with water, recoated for 10 min with 0.2% polybrene, and finally equilibrated with the running buffer for 5 min at 20 psi.

Since the control hIAPP kinetics may slightly vary depending on external features (such as the use of different hIAPP aliquots, colder or warmer room temperature), the evaluation of each hairpin was carried out simultaneously with the hIAPP control experiment to avoid any bias.

As explained above, we designed peptides derivatives mimicking acyclic β-hairpin structures selective for hIAPP. Our choice for the first SRE to link to the piperidine-pyrrolidine β-turn inducer, turned on the amyloidogenic sequence N₂₂FGAIL₂₇ because this region is strongly involved in hIAPP aggregation process and in the folding of hIAPP described in experimental and in silico models of monomers and aggregates found in the literature (Sumner Makin and Serpell, 2004; Kajava et al., 2005; Luca et al., 2007; Wiltzius et al., 2008; Bedrood et al., 2012; Liang et al., 2013; Wu and Shea, 2013; Tran and Ha-Duong, 2016; Cao et al., 2020; Röder et al., 2020). Furthermore, this sequence has been reported to be able to promote hIAPP fibrillization (Porat et al., 2004) and to inhibit its aggregation when inserted in a macrocycle peptide derivative (Buchanan et al., 2013). More recently, a series of hIAPP-derived 21-residues peptides, based on this hot segment FGAIL sequence, was designed by the group of Kapurniotu and reported as nanomolar inhibitors of hIAPP and Aβ140 self-assembly (Andreetto et al., 2015; Niu et al., 2020). Selecting the second SRE was trickier because the sequences facing N₂₂FGAIL₂₇ vary from one model to another being slightly shifted within the sequence A₁₃ to S₂₀. The selection of the complementary SRE in compounds 1 and 2 was based on models reported by Kajava et al. (2005), Andreetto et al. (2015) and Mirecka et al. (2016), where A₁₃NFLV₁₇ is facing N₂₂FGAIL₂₇. The only difference lies in the inversion of the SRE position with NFGAIL at the N-terminus in 1 or at the C-terminus in 2 (Figure 1). Compounds 3 and 4 (Figure 1) were based on models reported by Wiltzius et al. (2008), Liang et al. (2013), and Wu and Shea (2013) where F₁₅LVHSS₂₀ is facing N₂₂FGAIL₂₇. We kept five residues in each sequence, as pentapeptide sequences were sufficient in our hairpin mimics inhibiting Aβ_1–42 (Pellegrino et al., 2017). On the contrary to Aβ_1–42, hIAPP does not present acidic residue in the amyloidogenic region with which the terminal amine of the inhibitor would establish interesting ionic interactions. Thus, we introduced in compound 4 an N-acetyl group. The isolated SREs F₁₅LVHS₁₉ 5 and F₂₃GAIL₂₇ 6 (Figure 1) were also prepared to demonstrate the interest of inserting the SREs in β-hairpin mimics. The next step was to decrease the peptidic character in the β-hairpin mimics. We kept the sequences A₂₅IL₂₇ and F₁₅LV₁₇ that were present in compounds 1–3. The terminal amine was either free or protected by three different groups: Boc, acetyl or trifluoroacetyl (compounds 7–10). The free amine should increase the water solubility but is probably not mandatory for the interaction with hIAPP, as described above for 4. The trifluoroacetyl group was selected to examine how the presence of fluorine atoms could influence the activity. Fluorine possesses the unique capacity to increase the local hydrophobicity, to strengthen hydrogen bonds and to increase metabolic stability, that have been exploited in more than 300 approved drugs but not yet in peptide pharmaceuticals (Inoue et al., 2020). Fluorine has been also recently demonstrated as a useful probe for studying biological events by ¹⁹F-NMR spectroscopy (Dalvit and Vulpetti, 2019). The shorter compounds 11 and 12 were also evaluated to validate the interest of linking two peptide arms to the piperidine-pyrrolidine β-turn inducer. Finally, one α/aza/aza/pseudotripeptide, Phe-azaLys-azaVal-COCH₃ (abbreviated as FaLaV-COCH₃) mimicking the C-terminal arm FLV of 8, was inserted in compounds 13 and 14 and evaluated alone (compound 15). The introduction of one aza-amino acid in peptides has shown significant success in providing biologically active peptides (see the references cited in Tonali et al., 2020). However, to our knowledge, only scarce examples of biologically active peptide mimics having two consecutive aza-amino acids have been reported (Gante et al., 1995; Han et al., 1998). We ourselves introduced a diazaGly motif (Dufau et al., 2012) in inhibitors of HIV-1 protease dimerization. Here, we want to assess whether the diaza-amino acid motif can retain the selectivity of natural peptides due to side chains.

The preparation of the β-hairpin mimics 1–4 and of the SREs 5 and 6, was performed by solid phase peptide synthesis, using the Fmoc strategy on a rink amide resin, as previously reported by some of us (Pellegrino et al., 2017). The solution synthesis, outlined in Scheme 1, allowed the preparation of the target compounds 7–14 and start from scaffold 16 that was prepared in accordance with our published procedure (Pellegrino et al., 2014, 2017). To the N-terminus of 16, the three amino acids N-Fmoc-L-Leu-OH, N-Fmoc-L-Ile-OH, and N-Boc-L-Ala-OH were successively coupled to form the tripeptide AIL, in order to obtain the key intermediate 11. The coupling reagents HBTU/HOBt in the presence of collidine in dry DMF allowed the couplings in very satisfactory yields (80–96%, Scheme 1). The saponification of the methyl ester of 11, using NaOH 2 M in MeOH, afforded 19 in an excellent yield (96%). Its coupling with the tripeptide FLV-NH₂ 12 (synthesized in good yield, see experimental section), using COMU/Oxyma as coupling agent, in the presence of DIPEA in dry DMF, led to 7 in a very good yield (87%). Then, a classical acidic cleavage of the Boc moiety of 7 gave 8 in quantitative yield. N-acylation of 8 using acetic anhydride or trifluoroacetic anhydride, provided 9 and 10, respectively. The coupling between the intermediate 19 and the peptidomimetic arm FaLaVCOCH₃ 15, (whose synthesis is explained just below and in Scheme 2) resulted more difficult than in the case of the coupling with the peptide FLV-NH₂ 12. The desired compound 13 was obtained in satisfactory yield (59%), together with a compound resulting from the intramolecular cyclization between the carboxylic group and the NH of the sulphonamide of 19 (Supplementary Figure 28). The formation of the cyclized by-product only observed in the coupling with the di-aza-tripeptide could be due to its steric hindrance and/or lower reactivity. Finally, the acid cleavage of the Boc moiety of 13 afforded the final compound 14 in quantitative yield.

The preparation of the di-aza-tripeptide FaLaVCOCH₃ 15, reported in Scheme 2, required the combination of hydrazine chemistry and peptide synthesis. As recently reported by our group (Bizet et al., 2018), the introduction of two sequential aza-amino acids can be challenging because of the capricious reactivity of the alkyl-hydrazine. In our case, it was possible to prepare the di-aza-tripeptide arm 15 by elongating the peptidomimetic from the C-terminus to the N-terminus. Acetylation of Boc-2-isopropyl-hydrazine (20) prepared according to our reported procedure (Bizet et al., 2018), afforded 21 in a good yield (85%). The classical acidic cleavage of the Boc moiety of 21 gave 22. The activation of 22 using 4-nitrophenyl chloroformate, followed by the reaction with Boc-2-isobutyl-hydrazine led to the Boc-protected intermediate 23 in a very satisfactory yield (85%). The acidic cleavage of the Boc moiety of 23 followed by the coupling reaction of the intermediate 24 with N-Cbz-L-Phe-OH gave 25 in very good yield (80%). The cleavage of the Cbz moiety of 25 to afford the desired FaLaVCOCH₃ 15 was achieved employing McMurray’s hydrogenolysis (triethylsilane in the presence of Pd/C in MeOH (Mandal and McMurray, 2007).

In order to check if the conformational behavior of our new peptidomimetics was similar to the one of our previously reported acyclic β-hairpin structures bearing SREs based on Aβ_1–42 (Pellegrino et al., 2017; Tonali et al., 2018), we performed conformational studies by NMR on the most active inhibitors of hIAPP aggregation 3–4, 8–9, and 14, as assessed by ThT fluorescence assay (see below). The polar and protic solvent CD₃OH was used instead of water because of the low solubility in water of some compounds. Another important advantage of CD₃OH is the possibility to record NMR spectra at low temperatures (248–273 K range), thus allowing to better characterize slowly interconverting conformers. When signals were sufficiently resolved and disperse, vicinal ³J_HN–Hα coupling constants, H^α-HN ROE (Rotating-frame Overhauser Effect) correlations, temperature coefficient (Δδ_HN/ΔT) of the amide protons and ¹H_α and ¹³C_α chemical shift deviations (CSD), were examined to analyze backbone conformational propensities (see Supplementary Material for detailed data).

Compounds 3 and 4 bearing five residues sequences adopt similar β-hairpin like structure observed in our similar peptide compounds (Pellegrino et al., 2017) as confirmed by a good dispersion of the NH chemical shifts, sequential CHα_i/NH_i+1 ROEs for the extended peptide arms and characteristic ROE proximities in the beta-turn region and between the two strands (see Supplementary Material).

Conversely, compounds 8 and 9, bearing two shorter tripeptide arms, displayed the presence of at least two major conformers in dynamic equilibrium. Complete assignment of the ¹H and ¹³C signals was performed for the two conformers of compound 9 at 278 K (Supplementary Tables 6, 7). At this temperature, an equilibrium between two major conformers (ratio 1:1, 9a/9b) was observed. Large vicinal ³JNH-H_α constants (>8.0 Hz) were observed in both cases, suggesting the presence of dihedral angles typical of peptide segments in extended conformation. Several sequential CHα_i/NH_i+1 ROEs, indicating β-conformations, were found for both 9a and 9b conformers (Figure 2). The extended conformation of the two peptide arms was assumed by the positive difference between experimental Hα and HN chemical shift values and “random ones” (CSDs) (Supplementary Tables 8, 9). The very slight differences in values of ³JNH-Hα and of CSD let consider that the peptide arms were similarly extended in both conformers 9a and 9b. ROESY experiments confirmed the presence of the β-turn in both conformers (for details, see Supplementary Material). However, we detected inter-strand diagnostic ROEs only for 9b isomer (Figure 2), confirming spatial proximity between the two peptide arms. In particular, ROE contacts were visible between the Hα of Ile-2 and the aromatic ring of Phe-6 and the CH₃ of Leu-7, between the Hα of Leu-7 and the CH₃ of Ile-2 and, finally, between the side chains of Ile-2 and Leu-7 (Figure 2 right and Supplementary Figure 23). This difference in ROEs let us hypothesize a dynamic equilibrium between two different β-hairpin architectures, with only one of them allowing the spatial proximity between the peptide arms. This difference between the two conformers also results in a different ROE contact between the Hα of Leu-3 and the diastereotopic protons H2 or H6 of Pip-4. In 9a, Leu-3 Hα is closed to H6, while in 9b Leu-3 Hα provides a ROE with H2 (Figure 2 and Supplementary Figure 21).

We decided to perform a molecular dynamics (MD) study to investigate if a different torsion angle of the amide bond between Leu-3 and Pip-4 could explain the two different β-hairpin architectures observed experimentally.

200 ns-MD simulations of both conformers 9a and 9b were performed in explicit MeOH solvent at 280 K in order to be as close as possible to experimental conditions. In both initial structures, the torsion angle between the piperidine and proline rings was identical with proline Hα pointing toward piperidine H2 (torsion angle Cα-N-C3-C2 around −60°), as found in NMR analysis. The main difference between the two initial structures was the torsion angle around the amide bond between Leu-3 and Pip-4 (Figure 2): the Cα-C-N-C2 angle was equal to −170° and −5° in the initial structures 9a and 9b, respectively. The piperidine-proline local structure was maintained and no conformational transition of the corresponding initial torsion angle Cα-C-N-C2 between Leu-3 and Pip-4 occurred in MD simulations, for both conformers 9a and 9b (see section “Discussion” and Supplementary Figure 24). The computed ³J_HN–Hα coupling constants had lower values than NMR data, indicating that the two peptide arms were slightly less extended in MD simulations than in NMR experiments (see section “Discussion” and Supplementary Table 14). Interestingly, as observed in our NMR studies, significant differences in the spatial proximity of the two peptide arms were observed between modeled conformers 9a and 9b. Indeed, when analyzing the probability distribution of the distances between protons for which ROEs were detected (Figure 3), the distance between Ile-2 Hα and Phe-6 aromatic ring as well as that one between Ile-2 side chain CH₃ and Leu-7 Hα were clearly lower in 9b than in 9a. This indicates that the two peptide arms are closer in conformer 9b than in 9a (Figure 3). These results confirm that the conformation of the torsion angle Cα-C-N-C2 between Leu-3 and Pip-4 impacts the spatial proximity of the two peptide arms and the global β-hairpin architecture. When Leu-3 Hα is oriented toward piperidine H6 (9a), this promotes a separation of the two peptide arms, whereas when it is oriented toward piperidine H2 (9b), the two arms can approach one to the other.

This conformational analysis by NMR and molecular dynamics, performed on the acetylated hairpin mimic 9, allowed us to establish some general guidelines, useful for the structural characterization of similar compounds based on our β-turn inducer when NMR assignment was not as precise as for compound 9. In the case of the free N-terminal compound 8, it was not possible to perform a complete assignment of the two major conformers in equilibrium, even by lowering the temperature at 278 K. By increasing the temperature at 313 K, we could approach an approximately complete coalescence of the two conformers, allowing the attribution of an average conformation with a complete ¹H and ¹³C assignment (Supplementary Table 3). The good dispersion of the NH chemical shifts indicated the presence of a single and major conformation whose turn structure was confirmed by ROESY experiments. Several sequential CHα_i/NH_i+1 ROEs, large ³J_NH–Hα coupling constants and positive ¹H CSD values (Supplementary Table 4) for most of the amino acids (less remarkable for the more dynamic N-terminal Ala and C-terminal Val), suggested the extended conformation of the two peptide arms. Diagnostic inter-strand ROEs were observed between the NH of Leu-3 and the Hα and Hβ of Phe-6 and were in favor of a β-hairpin architecture (Supplementary Figures 13, 14). No one of the amide protons showed a temperature coefficient lower than −4.5 ppb K^–1 (Supplementary Table 5). However, Ile-2, Leu-3, and Leu-7 values fell within intermediate values, suggesting their partial involvement in hydrogen-bond. Taking together all these experimental results, it is possible to conclude that at 313 K, the average conformation of 8 adopts a partial β-hairpin conformation, with the N-terminal Ala residue not in contact with the C-terminal acetyl amide group, contact that would have allowed to form a more extended β-hairpin.

The ¹H and ¹³C assignment of compound 14 was tricky even performed on a spectrometer operating at 800 MHz and equipped with a TCI cryoprobe. A dispersion of NH chemical shifts and poor resolution of the signals indicated the presence of different conformers whatever the temperature from 278 to 313 K. Some signals and ³J_NH–Hα coupling constants of the natural amino acids could not be determined with certainty (Supplementary Table 15). The temperature dependence of amide proton chemical shifts of Ile-2 was intermediate (−4.5 ppb K^–1, Supplementary Table 16), suggesting its partial involvement in hydrogen-bond. Several sequential CHα_i/NH_i+1 ROEs and large and positive ¹H CSD values (Supplementary Table 17) for Ile, Leu, and Phe confirmed the extended conformation of the natural amino acids in the two arms (Supplementary Figure 30). No typical spatial proximities were found between H_α Pro-5 and Pip-4 and between the Pip-4 and NH Phe-6 to confirm the turn structure. Moreover, it can be affirmed that an equilibrium between two different β-hairpin architectures, one of which more “open” than the other, occurs even for compound 14, reflecting the same behavior as conformers 9a and 9b. A very weak inter-strand ROE was observed and attributed to Phe-6 Hδ and Leu-3 Hδ or Ile-2 Hγ whose chemical shifts were identical (Supplementary Figure 31). Overall, 14 seems to adopt a more flexible β-hairpin structure than the peptide analogs 8 and 9.