Local Control of the Cis–Trans Isomerization and Backbone Dihedral

Article pubs.acs.org/JPCB

Local Control of the Cis−Trans Isomerization and Backbone Dihedral Angles in Peptides Using Trifluoromethylated Pseudoprolines Debby Feytens,†,‡,§ Grégory Chaume,¶ Gérard Chassaing,†,‡,§ Solange Lavielle,†,‡,§ Thierry Brigaud,¶ Byung Jin Byun,⊥ Young Kee Kang,*,⊥ and Emeric Miclet*,†,‡,§ †

Laboratoire des BioMolécules, UPMC Paris 06, 4, Place Jussieu, 75005 Paris, France Laboratoire des BioMolécules, Departement de Chimie, Ecole Normale Superieure, 24, rue Lhomond, 75005 Paris, France § UMR 7203, FR 2569, 4, Place Jussieu, 75005 Paris, France ¶ Laboratoire SOSCO, Université de Cergy-Pontoise, EA 4505, 5 mail Gay Lussac, 95000 Cergy-Pontoise, France ⊥ Department of Chemistry, Chungbuk National University, Cheongju, Chungbuk 361-763, Republic of Korea ‡

S Supporting Information *

ABSTRACT: NMR studies and theoretical calculations have been performed on model peptides Ac-Ser(ΨPro)-NHMe, (S,S)Ac-Ser(ΨH,CF3Pro)-NHMe, and (R,S)Ac-Ser(ΨCF3,HPro)-NHMe. Their thermodynamic and kinetic features have been analyzed in chloroform, DMSO, and water, allowing a precise description of their conformational properties. We found that trifluoromethyl Cδ-substitutions of oxazolidine-based pseudoprolines can strongly influence the cis−trans rotational barriers with only moderate effects on the cis/trans population ratio. In CHCl3, the configuration of the CF3−Cδ entirely controls the ψdihedral angle, allowing the stabilization of γ-turn-like or PPI/PPII-like backbone conformations. Moreover, in water and DMSO, this Cδ-configuration can be used to efficiently constrain the ring puckering without affecting the cis/trans population ratio. Theoretical calculations have ascertained the electronic and geometric properties induced by the trifluoromethyl substituent and provided a rational understanding of the NMR observations.

■

INTRODUCTION The cyclic nature of the proline residue gives rise to some unique conformational properties of the peptide backbone. First, it induces a constraint by restricting the ϕ dihedral angle to values around −60°.1 Second, the Xaa−Pro peptide bond is subject to cis−trans isomerization characterized by an increased cis population (lower free energy difference ΔGtc) and an activation energy (ΔG⧧tc) that is low when compared to the other amino acids (Figure 1).2−4 Besides, the five-membered

ring of the Pro residue can adopt two distinct conformations (up-puckered or Cγ-exo and down-puckered or Cγ-endo)1 that are almost equally abundant in peptides5−7 and proteins.2,8,9 A variety of mimics and analogues have been designed in order to control the conformation of the peptide backbone and/or to alter the cis/trans ratios and the rotational barriers for cis−trans isomerization.10 In this context, Cδ-substituted prolines 1 and pseudoprolines (ΨPro) 2 have been shown to be very useful, as ΔGtc and ΔG⧧tc can be monitored depending on the nature of R1 and R2 as well as on the absolute configuration at Cδ.11−15 These molecules can be used to tailor cis/trans isomerization around the Xaa-ΨPro amide bond, offering a wide range of applications in peptide engineering. Theoretical calculations on Pro- and ΨPro-containing model peptides, such as Ac-Pro-NHMe, Ac-Oxa-NHMe (ΨPro Received: January 10, 2012 Revised: March 1, 2012 Published: March 8, 2012

δ

Figure 1. C -substituted prolines 1 and pseudoprolines 2. © 2012 American Chemical Society

4069

dx.doi.org/10.1021/jp300284u | J. Phys. Chem. B 2012, 116, 4069−4079

The Journal of Physical Chemistry B

Article

TROSY experiments.29 2D ROESY and NOESY experiments reflecting the spatial proximity between the CH3 acetyl group and the Hα or Hδ protons were used to assign the trans and cis conformations, respectively (mixing time from 300 ms to 1 s, typically collected as 1024 (t1) and 2048 (t2) time-domain matrices over a spectral width of 10 ppm, with 16 scans per t1 increment). 3JHα‑Hβ2 and 3JHα‑Hβ3 coupling constants used to determine the ring puckering were obtained from well-resolved 1D spectra or from CH2-TROSY experiments.30 Cis−trans isomerization rate constants were determined from the coalescence temperatures observed in 1D spectra (temperature studies between 233−323 K in CDCl3, 277−362 K in D2O:H2O 1:9, or 292−362 K in DMSO-d6).31,32 For each sample, typically three coalescence temperatures were determined on three couples of resonances, which were used for the calculation of independent rate constants. An additional method was tested for the measurements of rate constants, which is based on 2D EXSY experiments employing short mixing times (from 20 to 100 ms).14 A very good agreement with the coalescence temperature method was obtained. Rotational barriers were then calculated using the Eyring equation. Computational Methods. Chemical structures and torsional angles for the peptides 3, 4, and 5 are defined in Figure 2. All ab initio HF and hybrid density functional B3LYP calculations were carried out using the Gaussian 03 package.33 The 10 local minima (tAd, tCd, tFd, tAu, tCu, tFu, cAd, cFd, cAu, and cFu) and four transition states (ts1−ts4) for the Oxa peptide 3 optimized at the HF/6-31+G(d) level of theory in the gas phase and in solution (chloroform and water)21 were used to generate the initial structures for both the CF3-Oxa peptides 4 and 5 using the GaussView program.34 Then, all these initial structures were optimized at the same HF/631+G(d) level of theory in the gas phase and in solution. The first letter of each conformational letter code depicts the geometry of the ω′ prolyl peptide bond, either cis (c) or trans (t). The second informs about the (ϕ, ψ) backbone dihedral angles that correspond to the α-helical (A), γ-turn (C), or polyproline-like (F) regions of the Ramachandran diagram. The last letter depends on the χ1 angle of the oxazolidine ring and accounts for the up (u) or down (d) puckering. We employed the conductor-like polarizable continuum model (CPCM),35 implemented in the Gaussian 03 package,33 to compute solvation free energies (ΔGs) at the HF/631+G(d) level of theory with the UAKS cavities, which are the united atom topological model (UATM) radii optimized at the density functional PBE0/6-31G(d) level of theory.36 The solvation free energy is the sum of the electrostatic free energy and the nonelectrostatic energy terms. The latter is composed of the cavitation, dispersion, and repulsion energy terms. For CPCM-UAKS calculations, the default average areas of 0.2 Å2 for tesserae were used. The dielectric constants of chloroform and water used here are 4.9 and 78.4 at 25 °C, respectively. The structures of the transition states ts1 and ts2 are similar to the syn/exo structures with down and up puckerings, respectively, whereas those of ts3 and ts4 resemble the anti/exo structures with down and up puckerings, respectively, according to the definition of Fischer et al. in ref 37. Each transition state was checked by the intrinsic reaction coordinate (IRC) method38 whether it connects the reactants and products, that is, the trans and cis conformers. However, as in most cases, the IRC calculation did not step all the way to the minimum on either side of the path.39 Further optimizations were carried out

derived from Ser), and Ac-Thz-NHMe (ΨPro derived from Cys) have been reported in which ΔGtc and ΔG⧧tc as well as puckering transitions were calculated.15−22 When NMR data were available, theoretical calculations showed fair agreement with the observations gathered in solution: (i) replacement of Pro by ΨPro resulted in increased cis populations and decreased rotational barriers for cis−trans isomerization, (ii) presence of R1 substituents further decreased the rotational barriers, and (iii) presence of R2 substituents caused major increases of the cis populations. In the course of our investigations on the stereoselective synthesis of trifluoromethyl group containing amino acids (Tfm-AAs) and their incorporation into peptides,23 we have recently reported the synthesis of CF3-substituted pseudoprolines (CF3-ΨPro) as hydrolytically stable proline surrogates.24 Compared with other prolines or pseudoprolines substituents described in the literature, CF3 is unique being a bulky group,25 which displays strong inductive effects.26 Moreover, by analogy with Tfm-AAs and pseudopeptides,27 specific effects are expected from the incorporation of CF3-ΨPro into peptides such as a better stability toward proteases, an increase in lipophilicity, and a better affinity for lipid membranes.28 Herein, we establish the pronounced stereoelectronic influence of the trifluoromethyl Cδ-substituent, in both possible configurations, of the Ser-derived pseudoproline, i.e., (4S)-oxazolidine-4carboxylic acid (Oxa), on the thermodynamic and geometrical properties of model peptides. Both parameters were studied by NMR and theoretical calculations on the pseudopeptides: AcSer(ΨPro)-NHMe (Oxa peptide) (3), (2S,4S)Ac-Ser(Ψ H,CF3 Pro)-NHMe (2(S)-CF 3 -Oxa peptide) (4), and (2R,4S)Ac-Ser(ΨCF3,HPro)-NHMe (2(R)-CF3-Oxa peptide) (5), shown in Figure 2.

Figure 2. Model peptides 3, 4 and 5 used in this study.

■

MATERIALS AND METHODS Conformational Analysis by NMR. NMR experiments were recorded on a Bruker Avance III spectrometer operating at a 1H frequency of 500 MHz and equipped with a triple resonance, z-axis pulsed-field-gradient cryogenic probe head, optimized for 1H detection. Data were processed using TOPSPIN software using shifted sine-bell window functions and extensive zero-filling prior to Fourier transformation to yield high digital resolution. NMR samples were prepared by dissolving 2.5 mg of each compound in 600 μL of solvent (DMSO-d6; 0.03% TMS in CDCl3; 1% (v/v) DSS in D2O:H2O 1:9). Residual DMSO, TMS, or DSS signals were used to calibrate the chemical shift. 2D COSY and TOCSY experiments (mixing time of 80 ms, typically collected as 1024 (t1) and 2048 (t2) time-domain matrices over a spectral width of 10 ppm, with 16 scans per t1 increment) were recorded to obtain the complete proton assignments. Carbon assignments were obtained from heteronuclear 2D 1H−13C HSQC spectra using 512 (t1) and 1024 (t2) time-domain matrices over a spectral width of 10 ppm, with 32 scans per t1 increment or from CH24070

dx.doi.org/10.1021/jp300284u | J. Phys. Chem. B 2012, 116, 4069−4079

The Journal of Physical Chemistry B

Article

Table 1. Thermodynamic Properties of 3, 4, and 5 As Determined by NMR Spectroscopya solvent CDCl3

DMSO-d6

D2O:H2O 1:9

a

cis:trans ΔG°tcb ΔG ⧧tc ΔG ⧧ct cis:trans ΔG°tcb ΔG ⧧tc ΔG ⧧ct cis:trans ΔG°tcb ΔG ⧧tc ΔG ⧧ct

3

4

5

(27 ± 3):(73 ± 3) 0.59 16.74 ± 0.27 16.10 ± 0.27 (46 ± 2):(54 ± 2) 0.09 17.77 ± 0.25 17.66 ± 0.25 (34 ± 2):(66 ± 2) 0.39 18.41 ± 0.23 17.93 ± 0.23

(40 ± 2):(60 ± 2) 0.24 16.71 ± 0.26 16.43 ± 0.26 (66 ± 2):(34 ± 2) −0.39 16.77 ± 0.25 17.22 ± 0.25 (43 ± 2):(57 ± 2) 0.17 18.06 ± 0.27 17.88 ± 0.27

(24 ± 3):(76 ± 3) 0.68 14.77 ± 0.25 14.07 ± 0.25 (55 ± 2):(45 ± 2) −0.12 15.32 ± 0.25 15.46 ± 0.25 (45 ± 2):(55 ± 2) 0.12 15.63 ± 0.23 15.51 ± 0.23

All energies are given in kcal/mol. bFrom ΔG°tc = −RT ln Ktc, where Ktc = [cis]/[trans] at T = 298.15 K.

oxygen versus those resulting from the CF3 group (for synthesis details, see section S1 of the Supporting Information). Theoretical studies on 3 were previously reported, but to our knowledge no experimental data are available.17,21,22 NMR Analysis of the Cis−Trans Isomerization. The cis− trans equilibrium was studied by 1H NMR in CDCl3, DMSOd6, and D2O:H2O 1:9 solutions by recording spectra at different temperatures. In all solvents, two sets of 1H resonances were observed at the lowest temperatures for 3−5, clearly indicating the presence of two slowly exchanging conformations (Figures S1−S3 of the Supporting Information). 2D NMR experiments allowed us to unambiguously assign the cis (ω′ = 0°) and the trans (ω′ = 180°) conformers, the CH3 acetyl group being in close proximity to the Hα or Hδ, respectively. The cis/trans populations were determined by simple integration of the resonances and were found independent of the temperature in the range tested. Cis/trans ratios are reported in Table 1, together with the corresponding Gibbs energy ΔG°. In chloroform, introducing the (2R)-CF3 substituent at the Cδ-position of the Oxa ring (5) has virtually no influence on the cis population in contrast to the (2S)-CF3 substituent (4), which significantly stabilizes the cis content. In water and DMSO, increased cis contents are observed for all the three peptides when compared to chloroform, with median values of 43% and 55%, respectively. Peptide 3 incorporating the unsubstituted Oxa typically displays ∼10% lower cis content than 4 and 5 in these polar solvents. However, relatively high cis content is observed for 4 in DMSO (66%). The coalescence temperature method was used to determine the rotational barriers for cis−trans isomerization.31 The results are shown in Table 1. For a given molecule, ΔG⧧tc and ΔG⧧ct increase with the polarity of the solvent, as already reported for Ac-Pro-NHMe,19 and as expected, since polar solvents normally stabilize both ground states (partially charged) and destabilize the transition state. When compared to the unsubstituted ring in peptide 3, a CF3-substituent with the (2S) configuration does not alter ΔG⧧ct to a great extent, whereas the (2R) configuration at Cδ-position significantly decreases the rotational barrier by about 2 kcal/mol. H-Bond Study. In order to examine the presence of an intramolecular H-bond between the C-terminal amide NH and the carbonyl of the N-terminal acetyl group, the solvent dependences of the NH chemical shift were studied.45 The results are shown in Table 2. For peptides 3 and 5, the chemical shifts of the NH protons in the trans conformation are much less influenced by the solvent than the NH protons in the cis

starting from the reactants and products obtained by the IRC method to reach the two minima that the transition state connects. Vibrational frequencies were calculated for all stationary points at the HF/6-31+G(d) and CPCM HF/6-31+G(d) levels of theory in the gas phase and in solution, respectively, which were used to compute enthalpies and Gibbs free energies at 25 °C and 1 atm. The scale factor of 0.89 was chosen to reproduce experimental frequencies for the amide I band of Nmethylacetamide in Ar and N2 matrixes.40 In particular, each transition state was confirmed by checking whether it has one imaginary frequency after frequency calculations. The zeropoint energy correction and the thermal energy corrections were used to calculate the enthalpy (H) and entropy (S) of each conformation.41,42 Here, the ideal gas, rigid rotor, and harmonic oscillator approximations were used for the translational, rotational, and vibrational contributions to the Gibbs free energy, respectively. The B3LYP/6-311++G(d,p) singlepoint energies were calculated for all local minima and transition states of two peptides 4 and 5 located at the HF/ 6-31+G(d) level of theory in the gas phase and the CPCM HF/ 6-31+G(d) level of theory in solution. The B3LYP/6-311+ +G(d,p)//CPCM HF/6-31+G(d) level of theory provided the populations of backbone and prolyl peptide bond for the proline17 and pseudoproline21 peptides that are comparable with the observed values in chloroform and water. The bond overlap index and the atomic charges of the carbonyl C and the prolyl nitrogen of the C−N bond for the privileged trans and transition-state conformers of each of three peptides 3−5 in chloroform and water were calculated at the B3LYP/6-311++G(d,p) level of theory using the natural bond orbital (NBO) method43 implemented in the Gaussian 03 package.33 The Wiberg bond index44 obtained by the sums of squares of off-diagonal density matrix elements of the C−N bond can be used as a measure of a bond overlap. In addition, second-order perturbation energies calculated at the B3LYP/6311++G(d,p) level of theory, which reflect the strength of donor → acceptor (or bonding → antibonding) hyperconjugative interactions between NBOs,43b were used to analyze the relative stabilities of the down and up puckerings and the transition states for peptides 4 and 5.

■

RESULTS The CF3-Oxa-containing peptides 4 and 5 chosen for this study are displayed in Figure 2. The unsubstituted Oxa peptide 3 was included in order to distinguish the effects related to the ring 4071

dx.doi.org/10.1021/jp300284u | J. Phys. Chem. B 2012, 116, 4069−4079

The Journal of Physical Chemistry B

Article

Table 2. Solvent and Temperature Dependence of the NH Chemical Shift ΔδNH at 283 K (ppm) entry 3 3 4 4 5 5

cis trans cis trans cis trans

ΔδNH/ΔT (ppb/K)

CDCl3 → DMSO-d6

CDCl3 → water

DMSO-d6

water

1.92 0.86 2.53 2.01 1.93 1.01

2.21 1.19 2.66 2.34 2.14 1.35

−4.0 −3.6 −4.0 −4.2 −5.3 −4.8

−7.1 −7.0 −7.3 −7.5 −7.8 −7.9

conformation, as demonstrated by their ΔδNH values. This is indicative of the presence of an intramolecular H-bond in the trans conformations, which is absent in the cis conformations. In pseudotripeptide 4, the NH protons show a similar solvent dependence indicating the absence of an H-bond in either conformation. Additionally, the temperature influences on the NH chemical shifts were analyzed in polar solvents (Table 2).45 Temperature coefficients in water are of the same magnitude for all peptides in both cis and trans conformations (ca. −7.5 ppb/K), indicating that no intramolecular H-bonds are present at all. The H-bonds observed in the trans conformations of 3 and 5 in chloroform are clearly not strong enough to compete with water. In DMSO, all T−coefficients (ΔδNH/ΔT) were found lower than −3 ppb/K. This value has been used in DMSO to discriminate between solvent-shielded amide protons engaged in intramolecular H-bond (high T−coefficients) and amide protons of unstructured peptides (low T−coefficients).46 However, T−coefficients greater than −3 ppb/K are usually measured on cyclic peptides and this cutoff value appears inadequate for small-sized noncyclic peptides. Hence, we used the difference of the T−coefficients in the two peptide bond conformations as an indicator of the propensity to form Hbonds. Significant increases are observed in 3 (+0.4 ppb/K) and 5 (+0.5 ppb/K) when rotating the prolyl peptide bond from cis to trans conformation, suggesting that the H-bonds observed in CDCl3 for the 3 and 5 trans conformers are still weakly stabilized in DMSO. Puckering. The puckering of the five-membered ring in pseudoproline residues can be determined from the coupling constants between the α and β protons. As shown in Figure 3a, the α proton appears as a triplet (t) in the up-puckered form, and as a doublet of doublets (d × d) in the down-puckered form; Hα−Hβ vicinal coupling constants can be determined on the Hα resonances, provided that well-resolved 1D spectra are obtained. More generally, the precise determination of the vicinal couplings are reliably extracted from the Hβ resonances in the CH2-TROSY experiment since (i) the 13C dimension offers spectral dispersion and (ii) the geminal 2J coupling is removed.30 Such a spectrum is shown in Figure 3b, all extracted vicinal couplings being reported in Table 3. Note that for peptide 5 in CDCl3 and DMSO, no coupling constant can be precisely determined due to a faster exchange between the cis and trans conformations, which broadens the peaks. The results clearly indicate that for peptides 4 and 5 one puckering is found for both the cis and trans conformations, whatever the solvent. However, this puckering depends on the absolute configuration at Cδ, peptide 4 being observed as a down-puckered form whereas peptide 5 adopts the uppuckered conformation. For the unsubstituted peptide 3, a slightly different result is obtained. In CDCl3 and DMSO, 3 mainly presents the down-puckered form, but trans states are

Figure 3. (a) Newmann projections of the two possible ring puckerings, and (b) βCH2 region of the CH2-TROSY spectrum recorded on 5 in D2O, 293 K. Cross sections consist of two doublets, which enable the measurement of 3J(HαHβ3) and 3J(HαHβ2). For each conformer, vicinal coupling values have been reported in hertz.

characterized by higher 3J(HαHβ2) values (ca. 3.6 Hz). This is probably the consequence of a small population of up-puckered rings in fast exchange with the major down-puckered conformer. In water, such rapid exchange occurs for 3, being even more pronounced for the cis conformer. Theoretical Calculations. Preferred Conformations. The backbone torsion angles, the endocyclic torsion angles, and the thermodynamic properties of local minima and transition states for (2S)- and (2R)-CF3-Oxa peptides 4 and 5, calculated at the B3LYP/6-311++G(d,p)//CPCM HF/631+G(d) level of theory in chloroform and water, are listed in Tables S1 and S2 of the Supporting Information. All calculated data for the Oxa peptide 3 at the same level of theory were taken from ref 21. In addition, the preferred structures of local minima and transition states for peptides 4 and 5 optimized in chloroform and water are shown in Figures S4 and S5 of the Supporting Information, respectively, and the corresponding structures of peptide 3 in water are shown in Figure 5 of ref 21. The populations of all local minima for the three peptides in chloroform and water are listed in Table 4. All populations were obtained by electronic energies (ΔEe) in solution at 25 °C, since the Gibbs free energies of the (2S)-CF3 peptide 4 in chloroform and the (2R)-CF3 peptide 5 in water resulted in an overestimation of the cis populations (80.1% and 62.8%, respectively, from Tables S1 and S2 of the Supporting Information), compared to the NMR experimental values (40% and 45%, respectively, see Table 1). However, similar values of cis populations were obtained for peptide 3 by electronic energies and Gibbs free energies in chloroform and water. Backbone Populations. Using the distribution of populations reported in Table 4, the populations of the backbone conformations C, A, and F and the cis conformers can be calculated. The (2S)-CF3 substituent led to the major (over 90%) PP-like conformations (F), and the highest percent of cis peptide conformer, with ∼40% in chloroform and water. The (2R)-CF3 substituent led to backbone populations and cis peptide bond similar to those observed for the unsubstituted 4072

dx.doi.org/10.1021/jp300284u | J. Phys. Chem. B 2012, 116, 4069−4079

The Journal of Physical Chemistry B

Article

Table 3. Major Ring Puckerings Deduced from the Hα Multiplets and Vicinal Coupling Constants (in hertz) Measured on the Hβ2 and Hβ3 Resonances for 3, 4, and 5 3 cis

3 trans

4 cis

4 trans

5 cis

5 trans

d×d 2.0 7.1 d×d 3.1 7.0 t 3.8 5.3

d×d 3.5 7.0 d×d 3.7 7.2 d×d 3.6 7.2

d×d