3222
J. Phys. Chem. B 2007, 111, 3222-3235
Two-Dimensional Infrared Spectral Signatures of 310- and r-Helical Peptides Hiroaki Maekawa,† Claudio Toniolo,‡ Quirinus B. Broxterman,§ and Nien-Hui Ge*,† Department of Chemistry, UniVersity of California at IrVine, IrVine, California 92697-2025, Department of Chemistry, UniVersity of PadoVa, 35131 PadoVa, Italy, and AdVanced Synthesis and Catalysis, Life Sciences, DSM Research, P.O. Box 18, 6160 MD Geleen, The Netherlands ReceiVed: NoVember 12, 2006; In Final Form: January 18, 2007
Two-dimensional infrared (2D IR) spectra of CR-alkylated model octapeptides Z-(Aib)8-OtBu, Z-(Aib)5-LLeu-(Aib)2-OMe, and Z-[L-(RMeVal)]8-OtBu have been measured in the amide I region to acquire 2D spectral signatures characteristic of 310- and R-helical conformations. Phase-adjusted 2D absorptive spectra recorded with parallel polarizations are dominated by intense diagonal peaks, whereas 2D rephasing spectra obtained at the double-crossed polarization configuration reveal cross-peak patterns that are essential for structure determination. In CDCl3, all three peptides are of the 310-helix conformation and exhibit a doublet cross-peak pattern. In 1,1,1,3,3,3-hexafluoroisopropanol, Z-[L-(RMeVal)]8-OtBu undergoes slow acidolysis and 310-toR-helix transition. In the course of this conformational change, its 2D rephasing spectrum evolves from an elongated doublet, characteristic of a distorted 310-helix, to a multiple-peak pattern, after becoming an R-helix. The linear IR and 2D absorptive spectra are much less informative in discerning the structural changes. The experimental spectra are compared to simulations based on a vibrational exciton Hamiltonian model. The through-bond and through-space vibrational couplings are modeled by ab initio coupling maps and transition dipole interactions. The local amide I frequency is evaluated by a new approach that takes into account the effects of hydrogen-bond geometry and sites. The static diagonal and off-diagonal disorders are introduced into the Hamiltonian through statistical models to account for conformational fluctuations and inhomogeneous broadening. The sensitivity of cross-peak patterns to different helical conformations and the chain length dependence of the spectral features for short 310- and R-helices are discussed.
I. Introduction Three-dimensional protein structures formed by the folding of polypeptide chains are intimately related to important physiological functions and protein-mediated diseases such as amyloid diseases. Detailed elucidation of secondary structure formation, stability, and their structural and dynamic properties has been one of the main topics studied in chemistry and biology. Among various experimental techniques, infrared spectroscopy has been widely used for such investigations because of the sensitivity of vibrational transitions to protein structure. Of particular interest is the amide I mode, appearing in the 1600-1700 cm-1 region, that involves mainly the CdO stretching of the peptide unit. Various secondary structural motifs including helices, sheets, turns, and coils have been identified through empirical relationships with their different amide I frequency distributions.1,2 The amide I band is increasingly used as a secondary structure marker to follow the kinetics of protein denaturation and folding.3 However, it can be difficult to obtain detailed structural information from conventional linear IR spectra because of the broad line width and complications from spectral inhomogeneity. Recent achievements in femtosecond two-dimensional infrared (2D IR)4-33 have offered a new way to reveal rich structural information embedded in amide I vibrational spectra.4-29 The 2D IR approach can be thought of as a vibrational analogue of * Author to whom correspondence should be addressed. Phone: (949) 824-1263. Fax: (949) 824-8571. E-mail:
[email protected]. † University of California at Irvine. ‡ University of Padova. § DSM Research.
2D NMR but with a much higher time resolution. Structurerelated parameters, such as dihedral angles of the peptide backbone and distance and orientation between local amide I modes at different positions, have been extracted from observed 2D IR spectral patterns.14,19,23 Site-specific isotope editing has been applied to measure the coupling and correlation between the labeled vibrators, providing residue-specific information.21,22 Dual-frequency 2D IR techniques allow measurement of the anharmonic coupling between amide I and amide II or amide A modes.7,13 Nonequilibrium structural evolutions of peptide backbones triggered photochemically have also been investigated using 2D IR as a spectral probe.8 In this work we investigate 2D IR spectral signatures of 310and R-helices in the amide I region. While the R-helix is the most common secondary structure motif in proteins, approximately 10% of all helices adopt the 310-helix conformation with an average length of 3.3 residues.34,35 The primary differences between 310- and R-helices are the intramolecular hydrogen bonds. They are between the CdO group of the ith amino acid residue and the N-H group of the (i + 3) residue for the 310-helix and between the i and the (i+4) residues for the R-helix. Experimentally determined 310- and R-helix structures in proteins and peptides have average dihedral angles (φ, ψ) of (-57°, -30°) and (-63°, -42°), respectively,34,35 slightly distorted from (-49°, -26°) and (-57°, -47°) of the frequently used model structures.36 The 310-helix has been found to play important functional roles in several proteins.37-39 It has been observed that some transmembrane channel-forming antibiotics contain a high content of 310-helix with CR-dialkylated amino acid residues such as R-aminoisobutyric acid (Aib, Chart 1).40
10.1021/jp0674874 CCC: $37.00 © 2007 American Chemical Society Published on Web 03/08/2007
2D IR Spectral Signatures of Helical Peptides CHART 1: Molecular Structures of the Cr-Dialkylated Amino Acid Residues and the N-Terminal Protecting Groupa
a Aib, R-aminoisobutyric acid; (RMe)Val, CR-methylvaline; Z, benzyloxycarbonyl; NMTA, N-methyltrimethylacetamide.
Moreover, recent theoretical and experimental evidence suggests that the 310-helix is a possible intermediate in R-helix folding and melting.35,41-43 It is important to develop experimental techniques that are capable of investigating the interplay between the two helix types on picosecond time scales to fully understand helix formation processes and to refine coil-helix transition theory. Conventional spectroscopic techniques such as 2D NMR, electronic and vibrational circular dichroism (ECD and VCD) have been used to gain insight into the 310-helix conformation in the solution phase.35,44-47 For example, diagnostic ellipticity patterns in ECD spectra were experimentally observed for synthetic 310-helix model peptides, which are distinctly distinguished from those of the R-helix.44-47 Although these techniques are powerful, they do not have sufficient time resolution to study the role of the 310-helix during the earliest steps of the helix formation processes. Linear IR spectroscopy in the amide I region has been applied to discriminate between short 310and R-helices. The absorption band centered at 1648-1660 cm-1 was empirically assigned to the R-helix, and that at 16601670 cm-1 to the 310-helix.2 However, the amide I band peak frequency is easily influenced by several interfering factors, including helical length, types of amino acid residues, structural distortion, and solvent interactions. Thus, a sharp distinction between 310- and R-helices cannot be made according to the amide I maximum in linear IR spectra.48-50 To overcome the difficulties encountered in these traditional spectroscopic techniques, we investigate the sensitivity of 2D IR spectroscopy to the subtle differences between the 310- and R-helical structures. As the 310- and R-helical structures are different in their dihedral angles, pitch, rotation per residue, and intramolecular hydrogen-bond geometry, they will give rise to different vibrational couplings among the amide I vibrators along the peptide backbone and exhibit different 2D IR cross-peak patterns. Our strategy11 is to enhance the observation of crosspeak patterns by employing a double-crossed polarization configuration that can suppress overwhelmingly strong diagonal peaks.6,14 One advantage of this approach is that a cross-peak pattern can be obtained in a single measurement without the need of taking a difference spectrum between measurements taken with parallel and perpendicular polarization configurations.12,19,20,27 We report 2D IR spectral signatures on three terminally protected octapeptides (Chart 1), 1 Z-(Aib)8-OtBu (Z, benzyl-
J. Phys. Chem. B, Vol. 111, No. 12, 2007 3223 oxycarbonyl; OtBu, tert-butoxy), 2 Z-(Aib)5-L-Leu-(Aib)2-OMe (Leu, leucine), and 3 Z-[L-(RMe)Val]8-OtBu ((RMe)Val, CRmethylvaline) in several organic solvents. The crystal structures of 1-3 have average backbone dihedral angles close to those of short 310-helices in proteins.44,51-53 It was concluded from previous NMR and VCD measurements that all three peptides and the strictly related N-acetylated homo-octapeptide, Ac-[L(RMe)Val]8-OtBu (Ac, acetyl), form a stable two-turn 310-helix over their entire length in weakly polar solvents such as in CDCl3.44,45,54,55 It is, therefore, expected that similar amide I cross-peak patterns would be obtained for 1-3 regardless of different amino acid residues. Moreover, L-(RMe)Val homooctapeptides undergo a 310-to-R-helix transition following the slow acidolysis of the tert-butyl ester in 1,1,1,3,3,3-hexafluoroisopropanol (HFIP) over a time scale of days.11,45,46 This interesting feature allows one to measure 2D IR spectra of 310and R-helical conformations under the same conditions without complicating effects such as different solvents, types of amino acid residues, and lengths of polypeptide chains. The interpretation of experimentally observed spectral patterns for 310- and R-helices requires simulations. Previous theoretical studies have examined the helical structure sensitivity of 2D IR spectroscopy for a 15-residue peptide with a variety of 13CdO isotopic labels24 as well as a 20-residue peptide.25 In the latter study two dominating transitions are associated with the infrared-active A- and E1-modes as previously defined for an infinitely long helix.56-58 On the basis that a unit cell of the ideal 310-helix structure contains only three residues, it seems plausible that a simple classification into the A- and E1-modes would be valid for octapeptide 310-helices. However, this expectation needs to be tested. Moreover, the effect of the terminal edges becomes increasingly important as the helical length becomes shorter. The variation of the local amide I frequencies along the peptide backbone, depending on whether the vibrational modes are involved in intramolecular hydrogen bonding or exposed to solvent interactions, could strongly influence 2D spectral patterns. Therefore we need to model the effects of the hydrogen bonds, solvent interactions, and backbone fluctuations and examine how the amide I spectral pattern depends on the helix length. We simulate 2D IR spectra by calculating the third-order response of the one- and two-quantum vibrational eigenstates that are obtained by direct diagonalization of the vibrational exciton Hamiltonian.1,15-28,59 Elements of our model build on previous knowledge of the nonlinear response, vibrational couplings, and diagonal frequency shifts. We describe the through-bond nearest-neighbor couplings based on ab initio calculations60-62 and use transition dipole interactions to describe the through-space couplings between nonadjacent sites.1,59 To more quantitatively describe the diagonal frequencies, we devise a new approach that links the hydrogen-bond strength to the local amide I frequency shifts, taking into account the effects of hydrogen-bond length, angles, and sites. The inhomogeneous distributions of peptide structures are modeled through statistical variations of the backbone dihedral angles. The degree of delocalization of the exciton eigenstates is analyzed. We calculate 2D spectra as a function of chain length and investigate the extent that the eigenstates possess A- or E1-mode-like properties by examining their phase correlation,27 transition dipole amplitude and angle, and frequency ordering. This paper is organized as follows. In the next section, our 2D IR spectrometer and measurements are briefly described. Experimental results of Fourier transform (FT) IR and 2D IR spectra are shown in section III. We describe in section IV the
3224 J. Phys. Chem. B, Vol. 111, No. 12, 2007 theoretical approach based on the exciton model of the amide I modes and compare simulated linear and 2D IR spectra for the peptide with the ideal 310- and R-helix structures as well as with the structures determined by X-ray diffraction analysis and 2D NMR measurements. The 2D IR spectra of the 310-helices that are one turn shorter and longer than octapeptides are also calculated to elucidate the chain length dependence of the crosspeak pattern. The 2D IR spectral signatures of the 310- and R-helix conformations are discussed in section V. Conclusions are given in the final section. II. Experimental Section A detailed description of our 2D IR spectrometer has been given in previous reports from this laboratory.11,14 In brief, midIR pulses at approximately 6 µm (1667 cm-1, a spectral width of 160 cm-1, and a pulse duration of ∼100 fs) were generated by difference frequency mixing of near-IR pulses from a homebuilt optical parametric amplifier. Three IR pulses with almost equal pulse energies (∼300 nJ) and wavevectors of ka, kb and kc were arranged in box geometry and focused onto a sample cell. Pulse sequences of a-b-c and b-a-c are defined as rephasing (R) and nonrephasing (NR), respectively. The delay time between the first and the second incoming pulses is denoted as τ, and that between the second and third pulses as T. The zero origin of τ and T was determined by measuring the nonresonant response of the neat solvent. The third-order nonlinear signal induced in the phase-matching direction of -ka + kb + kc was collimated and combined with a local oscillator (LO) pulse. The delay time between the third pump pulse and the LO pulse is denoted as t. The polarization directions of the three incoming IR pulses (a, b, and c for the ka, kb, and kc pulses) and the detected nonlinear signal (d) were independently controlled and denoted as 〈a, b, c, d〉, where IR pulses propagate approximately along the X-axis of the laboratory fixed coordinates (X, Y, Z), and the Z-axis is defined as the zero for polarization angles on the Y-Z plane. Two different methods were employed to acquire 2D IR spectra: heterodyne detection in the time domain and spectral interferometry in the frequency domain. For the heterodyne detection of the IR echo signal, the combined signal and LO pulse were sent to a single channel mercury cadmium telluride (MCT) detector. Interference between the solvent nonresonant signal and the LO pulse was measured to decide the zero origin of t. By chopping one of the pump pulses at 500 Hz, one-half of the repetition rate of the IR pulse generation, the LO pulse intensity was subtracted experimentally. The interferometric signal was recorded by scanning τ and t from 0 to ∼3 ps with a 9 fs step at the fixed delay time of T ) 0. Absolute magnitude 2D IR spectra were obtained by 2D Fourier transformation of the collected data matrix along τ and t whose conjugated frequencies are ωτ and ωt, respectively. In spectral interferometry measurements, the combined LO pulse and nonlinear signals were dispersed by a monochromator and detected by a 64-channel MCT array detector. The frequency resolution is approximately 2 cm-1/pixel at 6 µm. Wavelength calibration at each pixel was achieved by measuring the rovibrational absorption lines of water vapor. To determine the zero origin of the delay time t, the spectral interferogram of the nonresonant solvent signal was measured at a long t (about 2 ps). Fourier transform of the interferogram exhibits a peak at t in the time domain.63 The translation stage for the LO pulse was moved back to the zero origin of t based on this determination before measurements of sample solutions were performed. Data matrices of the 〈Z, Z, Z, Z〉 polarization
Maekawa et al. component were obtained by scanning τ from 0 to ∼3 ps with a 4 fs time step. The delay time T was fixed at 300 fs to reduce the nonresonant contributions of the solvent. After calibration of pixel-to-pixel responsivity dependence and scaling the intensity with a Jacobian factor of λ2 to convert from wavelength (λ) to wavenumber (ωt), 2D IR spectra were obtained by Fourier transformation of the data matrices along τ to give the conjugated frequency ωτ. The spectra obtained from R and NR sequences were summed to remove phase skewing and obtain a 2D absorptive spectrum.30,64 To ensure proper phase adjustment of the absorptive spectrum, its projection onto the ωt axis is compared with a dispersed IR pump-probe spectrum collected immediately after measurements of the R and NR spectrum. The dispersed pump-probe spectrum was measured at the same delay time T with the same polarization configuration by blocking the ka pulse and using the kc and kb pulses as the pump and probe pulses, respectively. The probe pulse was attenuated by a factor of 10. Linear IR spectra of the peptide solutions were measured with a purged FT IR spectrometer (Perkin-Elmer, Spectrum 2000). Each spectrum was collected with a 4 cm-1 resolution and averaged over 64 scans. Background spectra of the neat solvents were measured under the same conditions and subtracted from the peptide solution spectra. The 310-helix octapeptides 1-3 were synthesized and characterized as described previously.44,65,66 CDCl3 (Cambridge Isotope Laboratories, 99.96% D) solutions of 1 and 2 were held in a CaF2 sample cell with an optical length of 250 µm. To prevent peptide aggregation, the concentration was set to approximately 5 mM, leading to a maximum optical density of 0.2-0.3 for the amide I band. To prepare a solution of peptide 3 with a stable 310-helix conformation, it was dissolved in CDCl3. The 310-to-R-helix transition was induced in HFIP (Acros Organics, 99.5%) in parallel experiments. The concentrations were approximately 10 mM so that the optical density of the amide I band was approximately 0.3 in a 100-µm-thick sample cell. These solvents were used without further purification. All of the measurements were executed at ambient temperature (20 °C). III. Results A. 310-Helix of Aib-Rich Octapeptides (1 and 2). Figure 1a shows FT IR spectra of the Aib-rich octapeptides 1 and 2 in CDCl3. The amide I bands of 1 and 2 peak at 1666 and 1664 cm-1, respectively, and their line shapes are almost the same with a full width at half-maximum (fwhm) of 32 cm-1. The peak frequencies for these bands agree with those reported previously.44,55,67 A clear difference between 1 and 2 is found in the region above 1700 cm-1, where the urethane and ester CdO bands are observed. The CdO band of the methyl ester at 1735 cm-1 is separated from the urethane CdO band by 20 cm-1; however, the CdO band of the tert-butyl ester has a lower fundamental frequency, and hence it overlaps with the urethane CdO band. Figures 2a and 2b show the 2D IR absorptive spectra of 1 and 2 measured under the 〈Z, Z, Z, Z〉 polarization configuration with T ) 300 fs. The projection of the phase-adjusted 2D absorptive spectrum onto the ωt axis and the dispersed IR pump-probe spectrum collected at a delay time of 300 fs are also plotted in the panel above each 2D spectrum, illustrating the quality of the phase adjustment. In these 2D spectra, two pairs of positive- and negative-going peaks are observed. The peak maxima along the ωt axis are located at the same frequencies as those in the FT IR spectra. As the spectral width
2D IR Spectral Signatures of Helical Peptides
Figure 1. FT IR spectra of octapeptides 1-3 in CDCl3 and HFIP: (a) 1 (red), 2 (blue), and 3 (green) in CDCl3; (b) 3 in HFIP measured immediately (red) and 34 days (blue) after sample preparation. The background spectrum of the neat solvent is subtracted, and each spectrum is normalized by the peak absorbance of the amide I band at approximately 1660 cm-1.
Figure 2. Two-dimensional absorptive spectra measured with the 〈Z, Z, Z, Z〉 polarization configuration at T ) 300 fs: (a) 1 in CDCl3; (b) 2 in CDCl3; (c) 3 in CDCl3; (d) 3 in HFIP measured 34 days after sample preparation. All spectra are normalized by the maximum intensity of the positive peak. Equally spaced red and blue 12 contour lines are drawn at positive and negative values, respectively. The dotted line is drawn along the diagonal to direct the eye. The projection onto the ωt axis of the phase-adjusted 2D correlation spectrum (blue) is also shown in the panel above each 2D spectrum together with the dispersed pump-probe spectrum (red) measured with the parallel polarization configuration at a delay time of 300 fs between the pump and probe pulses.
of the IR pulse in our experiments is wide enough to excite not only the amide I but also the terminal CdO bands, we can observe both of their contributions in the 2D spectra. Because of the anharmonicity of these vibrators, the negative peaks are shifted to lower frequencies along the ωt axis and separated from the positive peaks at the diagonal position (ωτ ) ωt). The most important information on peptide structure is contained in the vibrational couplings between the amide I modes of the different peptide units, which generate cross-peaks in 2D spectra. However, as shown in Figure 2, diagonal peaks are usually much more intense than cross-peaks and will veil cross-peaks if they are close in frequencies. One way to remove the strong diagonal peaks and observe only the cross-peaks is to make use of their different dependencies on polarization
J. Phys. Chem. B, Vol. 111, No. 12, 2007 3225 configurations. It has been shown theoretically and experimentally that the 〈π/4, -π/4, Y, Z〉 polarization configuration can suppress the diagonal peaks.6,14 To check the extent of diagonal peak suppression, we measured the absolute 2D IR spectrum of N-methyltrimethylacetamide (NMTA), a model compound of Aib, in CDCl3 under 〈Z, Z, Z, Z〉 and 〈π/4, -π/4, Y, Z〉. It was found that the diagonal peaks were suppressed to
