2.1 Peptide synthesis

Automatic peptide synthesis was achieved as described elsewhere [1,5]. The H–VPPPVPPRRR–OH peptide (V10R) was obtained from a Boc–Arg(Tos)–OH preloaded PAM resin (0.1 mmol): RP-HPLC, t_R = 10.3 min (30 min gradient from 9 to 48% CH₃CN). MALDI-Tof (M + H⁺) 1170.7 (calcd 1169.7). The N-acetylated and C-carboxamide V10R (Ac–V10R–NH₂) peptide was obtained from a MBHA resin. After removal of the last Boc-protecting group, the amino terminal was acetylated by an excess of acetic anhydride. RP-HPLC, t_R = 15.9 min (30 min gradient 6 to 30% CH₃CN). MALDI-Tof (M + H⁺) 1211.9 (calcd 1210.7).

2.2 Circular dichroism

CD spectra were acquired on a Jobin Yvon CD6 dichroism spectropolarimeter at 5 °C using a 1-mm pathlength quartz cell. Spectra were recorded from 190 to 260 nm and were averaged over four scans after baseline correction. Samples containing 2.0–7.4 M guanidinium hydrochloride (Merck) were obtained by adequate additions of 8.0 M concentrated solution. Samples containing 2.0–5.5 M calcium chloride (Sigma–Aldrich) were obtained by adequate additions of 6.0 M concentrated solution. For each sample, final V10R peptide concentration was fixed to 300 μM in a 2 mM phosphate buffer, pH 7.

2.3 NMR spectroscopy

NMR samples were prepared by dissolving lyophilized peptides (V10R or Ac–V10R–NH₂) to final concentrations of 2 mM in 550 μL phosphate buffer (20 mM, pH 6.7 in 10% or 100% D₂O). Sodium 2,2-dimethyl-2-silapentane-5-sulfonate-3,4,5-d₆ at 0.5 mM was used as an internal reference for ¹H chemical shift calibration. NMR experiments were recorded at 25 °C on a DMX Bruker Avance spectrometer operating at a ¹H frequency of 500 MHz and equipped with a room-temperature, triple-resonance, z-axis pulsed-field-gradient probehead, optimized for ¹H detection. Proton assignments were obtained from the analysis of 2D total correlation spectroscopy (TOCSY) experiments [6] using a 65 ms DIPSI-2 mixing time [7], 2D waterflipback nuclear Overhauser enhancement spectroscopy (NOESY) [8] and 2D rotating frame Overhauser effect spectroscopy (ROESY) [9] experiments (250 ms mixing time). Solvent suppression was achieved with a WATERGATE sequence [10]. Two-dimensional experiments were typically collected as 512 (t1) and 2048 (t2) time-domain matrices over a spectral width of 11 ppm, with 32 scans per t1 increment. Carbon assignment was deduced from proton assignment using heteronuclear 2D ¹H–¹³C HSQC spectra [11].

Data were processed with NMRPipe software running on a SGI O2 workstation [12]. Shifted sine-bell window functions were applied in both indirect and direct detected dimensions and extensive zero filling prior to Fourier transformation was used to yield high digital resolution. Spectra were analyzed using CARA software [13].

2.4 Structure calculation

Simulated annealing (SA) calculations were carried out with the program DYNAMO 2.1 [14] and consisted of three stages. The first stage comprised of an initialization period of 1000 steps (3 fs long) of molecular dynamics at 4000 K and very tight temperature control. In the second stage, the coordinates were allowed to cook at 4000 K with lose temperature control for 4000 steps. In the final stage, the structure was annealed by slowly reducing the temperature from 4000 to 0 K over the space of 20,000 molecular dynamic steps (3 fs long). In the first two stages, force constants were set as follows: 1000 kcal mol⁻¹ Å⁻², 250 kcal mol⁻¹ rad⁻², 50 kcal mol⁻¹ rad⁻², 10 kcal mol⁻¹ rad⁻², 2 kcal mol⁻¹ Å⁻² for bonds, angles, impropers, torsions and NOE constraints, respectively. No Van der Waals interactions were operative. During the final stage, bond force constant was maintained at 1000 kcal mol⁻¹ Å⁻², torsions were ramped from 10 to 200 kcal mol⁻¹ rad⁻², both angles and impropers to 500 kcal mol⁻¹ rad⁻², NOE term was increased exponentially from 2 to 30 kcal mol⁻¹ Å⁻². The Van der Waals interactions and J-coupling restraints were slowly introduced during cooling to reach the final values of 4 kcal mol⁻¹ Å⁻² and 1 kcal mol⁻¹ rad⁻², respectively. Six ϕ- and five ψ-torsions restraints have been predicted with good accuracy (ca. ±10°) using the program TALOS [15] and have been included in structure calculation. Other NMR constraints consisted of 16 vicinal J-coupling constants measured from 1D and/or 2D spectra and 62 NOE derived from NOESY and ROESY experiments. Hundred structures have been calculated starting from an extended fold. Typical final energies (kcal mol⁻¹) were (<0.2); (3.8 ± 0.3); (4.5 ± 0.5); (3.2 ± 0.4); (22.7 ± 1.5); (4.7 ± 0.5) and (14.9 ± 1.3) for torsion, J-coupling, bond, impropers, angle, NOE and Van der Waals force constants, respectively.

3.1 Relative stability of conformers

Since V10R contains five prolines, 2⁵ different conformers may be in slow exchange due to the cis–trans proline isomerisation. Then, a preliminary 1D NMR analysis has been performed to detect the number of V10R conformers present in solution. Four signals are expected per conformer, the 10 residues peptide containing five proline and the first residue amine protons being not visible in the spectrum because of fast exchange with water. Surprisingly, the ¹H amide region contained only few peaks, including four strong and four weak doublets (Fig. 1). One can directly conclude from this observation that among the possible 32 conformers only two were present in solution (conformers A and B). Moreover, signal integration gave a ratio of 90:10 for conformers A and B, respectively.

The same result was deduced from close inspection of 2D ¹³C–¹H HSQC spectra. In addition to two strong correlation peaks in the C^α–H^α valine region assigned to V1 and V5, one minor correlation (9%) was observed for B, suggesting that one of the two valine residues was sensitive to a slow exchange between two V10R conformers (Fig. 2a). The presence of this exchange was also confirmed in the C^γ–H^γ methyl region where two weak additional methyl correlations were observed (Fig. 2b). It is noteworthy that one of the two valine residues was not influenced by the conformational change of V10R, probably because proline isomerisation took place far in the sequence and did not induce chemical shifts perturbation.

Stability of the two conformers was also assessed as a function of temperature. Vicinal J3HN–Hα were very stable when raising the temperature from 5 to 65 °C (ca. < 0.01 Hz/°C, data not shown). Moreover, the number of conformers did not increase at high temperature as judged by analysis of proton 1D or 2D ¹³C–¹H spectra, and the ratio between the conformers A and B being also about 90:10 at 65 °C. No exchange peak between the two V10R forms was detected in NOESY spectra, even using long mixing time delays at high temperature [18]. Altogether, our data suggested that whatever the temperature (5–65 °C), two V10R conformers in a stable ratio of 91:9 were interconverting in the slow exchange regime.

3.2 Conformers properties

Complete and partial ¹H and ¹³C assignments have been obtained from homo- and heteronuclear 2D spectra for conformers A and B, respectively. Proline ¹³C^β and ¹³C^γ chemical shifts give direct information about Xaa–Pro peptide bond conformations [19]. Indeed, the C^β–C^γ correlation pattern of conformer A (Fig. 3) revealed that Pro2, Pro3, Pro4, Pro6 and Pro7 belong unambiguously to the trans region plot. This latter conformation was confirmed by NOESY spectra since strong H^α_Xaai−1–H^δ_Proi NOE were observed for these residues. Interestingly, the C^β–C^γ correlation plot of conformer B placed its Pro6 residue in the cis region. This peptide bond conformation was further supported by the NOESY spectra where a strong H^α_Val5–H^α_Pro6 cross peak was observed (Fig. 4). Thus, the five prolines of conformer A were found in the trans conformation, a feature compatible with a PPII fold, whereas the Pro6 residue of conformer B was in the cis form. This provides explanation why the immediately preceding Val5 experienced different chemical shifts for each conformer. In contrast, chemical shifts of other residues appeared unchanged or very close for conformers A and B, suggesting that Pro6 conformational state is the only structural difference between A and B.

3.3 Influence of the N-termination on peptide structure

The five V10R prolines can be classified into three groups: (i) prolines preceded by a non-prolyl residue (Pro2 and Pro6), (ii) prolines followed by a non-prolyl residue (Pro4 and Pro7) and (iii) prolines surrounded by two prolines (Pro3). One can observe that Pro6, the only proline experiencing cis–trans isomerisation, belongs to the group (i). However, the Val1–Pro2 peptide bond was >99% trans, as estimated from ¹³C–¹H spectra (Fig. 2a and b). To assess whether the N-termination of Val1 could influence Pro2 isomerisation, Ac–V10R–NH₂ was synthesised. Interestingly, the presence of the N-terminal acetyl group restored the ability of Pro2 to undergo the two conformational states. Indeed, secondary signals were observed for Val1 of Ac–V10R–NH₂ in both the C^α–H^α and C^γ–H^γ regions (Fig. 2c and d, respectively). Precise integration of strong and weak C^α–H^α, C^γ1–H^γ1 and C^γ2–H^γ2 correlation peaks gave 88% of trans conformation for the Val1–Pro2 peptide bond in Ac–V10R–NH₂, whereas the trans Val5–Pro6 peptide bond remained unchanged at 91%.

5.1 Structure calculation

Structure determination of conformer A has been carried out by NMR spectroscopy. Constraints consisted of a total of 62 NOE, 11 torsions restraints, 4 J3HN–Hα and 12 J3Hα–Hβ vicinal couplings. Among the 100 calculated structures, the best 10 are presented in Fig. 6. The first Val1–Pro7 segment adopted a left-handed helix conformation with three residues per turn. Proline torsion angles were closely restricted to the canonical PPII ϕ and ψ values (−67 ± 7° and 150 ± 12°, respectively). Interestingly, proline residues belonging to group (ii) (Pro4 and Pro7) had a slight 5–10° weaker ψ angle than those followed by prolines. Since Val1 (ψ ∼ 148°) and Val5 (ϕ ∼ −85° and ψ ∼ 132°) were also located in the PPII region of the Ramachandran plot, the overall fold of the Val1–Pro7 segment was a regular PPII helix, characterised by a very low RMSD (RMSD = 0.35 Å for backbone atoms and RMSD = 1.67 Å for all atoms taken from the 10 best structures). In contrast, the three C-terminal Arg residues were poorly structured, as a result of the lack of NOE constraints.

5.2 Comparison of the free and Grb2-bound structures

Structure calculations unambiguously led to an extended PPII structure for conformer A, which accounts for 91% in solution. Alignment with the structure obtained for the Grb2-bound peptide [2] gave a small RMSD value of 0.96 Å for backbone atoms and a RMSD of 1.65 Å for all atoms belonging to residues comprised between Val1 and Pro7. Close analysis of dihedral angles of this region revealed same features for both structures, suggesting that no major conformational change occurs when conformer A of V10R binds to Grb2 within this segment. In contrast, the basic C-terminal region of V10R must be stabilised upon binding since Arg8 directly interacts with an aspartate of the protein (Asp33 and Asp190 for the Grb2 N-terminal and C-terminal SH3 domains, respectively).