11628
J. Phys. Chem. B 2010, 114, 11628–11637
Comparison of Isotopic Substitution Methods for Equilibrium and T-Jump Infrared Studies of β-Hairpin Peptide Conformation Karin Hauser,*,†,‡ Oliver Ridderbusch,‡ Anjan Roy,§ Alexandra Hellerbach,† Rong Huang,§,| and Timothy A. Keiderling*,§ Institute of Physics, RWTH Aachen UniVersity, Sommerfeldstr. 14, 52074 Aachen, Germany, Institute of Biophysics, Goethe-UniVersity Frankfurt, Max-Von-Laue-Str. 1, 60438 Frankfurt, Germany, and Department of Chemistry, UniVersity of Illinois at Chicago, 845 West Taylor Street, Chicago, Illinois 60607-7061 ReceiVed: March 29, 2010; ReVised Manuscript ReceiVed: July 26, 2010
Laser induced temperature jump (T-jump) relaxation kinetics were measured with infrared absorbance (IR) detection for a set of β-hairpin peptides, related to the Trpzip2 hairpin, but containing single isotopic labels, 13 C on the amide CdO of selected residues both in the center of the strands and at the terminal regions of the hairpin. Variations in the behavior of single labeled peptides are compared to those previously reported for double labeled variants. Although single labels do not result in spectral intensity enhancement, as seen for cross-strand labeling, the IR frequency shifts are still diagnostic of hairpin unfolding. If CdO’s in the β-strand portion of the hairpin (between the Trp residues) are labeled, the dynamic behavior of the local modes is similar to the results obtained with double labels in terms of relaxation time and activation energy and closely tracks the kinetics of the β-strand components. This implies that either property, local secondary structure (change of φ,ψ), or cross-strand coupling enabled by strand formation and H-bonding relaxes with the same kinetic mechanism. Single labeled residues on the terminal positions have a different behavior and are less able to be detected due to overlap with the 12C components, in contrast to double labels involving these positions, which are enhanced due to coupling. DFT-based spectral simulations that use the NMR structure of Trpzip2C indicate that the single labeled peptides should have roughly equivalent 12C bands but the 13C mode frequencies will vary with sequence position. Effective solvent corrections using COSMO yield significant changes in the frequencies but not in the relative isotope shifts obtained in our calculated spectra. Sequence positional dependence of labels is shown to be more discriminatory for kinetics changes than for thermodynamic variations. Introduction Infrared (IR) absorbance is a powerful and versatile technique for sensing peptide secondary structure, but due to its resolution limitations, only average backbone conformational data can normally be determined. However, if peptides are selectively isotope labeled, certain IR bands become sensitive to sitespecific structural aspects of the peptide because of the vibrational coupling between the labeled residues.1-11 If the local mode is sufficiently sensitive to variation in the secondary structure, then its frequency shifts will be diagnostic of the site conformation, independent of coupling to other modes. These effects can be modeled using quantum mechanical (QM) force fields (FF) as well as empirical methods, as has been demonstrated by a number of applications.1,3,5,12-16 The key to utilizing any structural technique is reliable detection of some physical property that has a dependence on the structure, one whose response will reflect alterations in the structure, or some other aspect of the molecular system of interest. Through-space and through-bond coupling of local modes in residues generates frequency shifts that contribute strongly to characteristic IR and Raman frequencies and * To whom correspondence should be addressed. E-mail: hauser@ physik.rwth-aachen.de (K.H.);
[email protected] (T.A.K). † RWTH Aachen University. ‡ Goethe-University Frankfurt. § University of Illinois at Chicago. | Current address: Department of Pathology and Laboratory Medicine, University of Cincinnati, Cincinnati, OH 45237.
intensities, which are normally used for structural diagnoses. They also affect band shapes, particularly such as those used for vibrational circular dichroism (VCD) and for nonlinear, 2DIR, vibrational analyses of structure. Hence, it is natural to choose to label structurally close-lying sites in peptide sequences to enhance the impact that their coupling has on the spectra. Observation of changes in coupling during folding processes can be diagnostic of variations in local structure. NMR studies have similarly used magnetic dipole coupling between close lying labeled sites to evaluate structures, particularly for studies of aggregates.17-19 However, other labeling schemes might also be useful, and have been utilized in previous IR studies.2,6,9,16 In particular, it is important to ask what kind of information can be gained if only a single label is incorporated, since its location would be clearer than for multiple labels. One must also investigate from where any single label structural sensitivity arises. These sense local geometry or at least sequence through what might be termed the diagonal force-field (FF) components associated with that site and are not easy to predict, since they are sensitive to residue type and local conformation, disorder, and solvation. By contrast, the off-diagonal FF terms (that correspond to coupling between labeled residues) are relatively insensitive to these perturbations.12,20 The labeled residues can also couple to the other parts of the molecule (unlabeled residues) via other off-diagonal FF interactions, although the effect of such coupling
10.1021/jp1028245 2010 American Chemical Society Published on Web 08/13/2010
Comparison of Isotopic Substitution Methods on isotopically shifted bands would be reduced by the differences in diagonal frequencies. This paper provides example spectral studies using a model β-hairpin peptide that contrast use of either single or double (cross-strand) isotope labels at different locations to enable a comparison of the results of monitoring thermal variation with both equilibrium and dynamic (temperature jump) IR spectroscopy. Quantum mechanical (DFT-based) spectral simulations were additionally performed to aid in detailed band interpretation and to gain some sense of local (diagonal) vs coupling (offdiagonal) effects on the observed spectral band positions and intensities. The simplest β-sheet unit is a β-hairpin, in which two antiparallel β-strands are connected by a reverse turn (usually 2-6 residues). Among the many model β-hairpins that have been studied, the tryptophan zipper (or Trpzip) structures, which have four tryptophan residues forming a stable hydrophobic cluster in a sequence of 12-16 residues, have one of the most stable β-hairpin structures, particularly among those example hairpins with a tight turn. Consequently, Trpzip’s have been the focus of a number of experimental and theoretical studies.9,16,21-32 These hairpin peptides are monomeric, are watersoluble, and have significant thermal stability in aqueous solution, due to cross-strand hydrophobic interactions between the tryptophan residues.8,21,32-40 Trpzip2 has been described as the smallest peptide sequence with proteinic residues that has a tertiary structure without aid from disulfide bonds or metal binding.21 We have carried out a number of studies of Trpzip peptides using isotopic labeling and site mutation to better understand the forces that favor hairpin folding and the mechanisms that lead to structure formation.8,13,31,32,35 In this paper, we will focus on the Trpzip2C hairpin (TZ2C, AWAWENGKWAWK-NH2, where NG indicates the Asn-Gly residues that form a tight β-turn), which is a slight modification of the original Trpzip2 sequence of Cochran et al.21 Substitution of Ala for residues 1, 3, and 10 was done for ease of labeling, and TZ2C has (as we have shown by complete NMR structural studies) the same hairpin conformation as Trpzip2.8 The well-studied Trpzip2 model system was chosen due to the fact that extensive structural and extensive thermodynamic and kinetic data are already available, making it a good test system for analyses of substitution methods. Our purpose in this study is to compare the equilibrium thermodynamic and temperature jump (T-jump) kinetic analyses possible with hairpins having one isolated vs two coupled labels. Our results show each approach to have advantages. Methods Peptide Syntheses. These β-hairpin peptides were initially synthesized at the Protein Research Facility of the UIC Research Resources Center, using standard solid-phase (FMOC) peptide synthesis protocols. Labeled Ala (L-Alanine-N-Fmoc, 1-13C, CLM-116) was purchased from Cambridge Isotope Laboratories, Inc. Samples were HPLC purified, and their identities were confirmed by MALDI-MS. Table 1 summarizes the various labeling positions in the peptide sequence and the corresponding abbreviations that will be used in this paper, and Scheme 1 illustrates their relative positions with respect to the cross-strand H-bonds. For IR experiments, the purified peptides were lyophilized against 0.1 M DCl in D2O to remove residual trifluoroacetic acid (TFA), whose absorbance (∼1672 cm-1) can interfere with IR measurements. This also results in H/D exchange of the N-H
J. Phys. Chem. B, Vol. 114, No. 35, 2010 11629 TABLE 1: Definition of the Trpzip Sequences and Labeling Positions in the Peptide Sequence along with Their Abbreviations Used in This Paper abbreviation
peptide sequencea
TZ2C A1 A3 A10 A1A10 A3A10
AWAWENGKWAWK-NH2 AWAWENGKWAWK-NH2 AWAWENGKWAWK-NH2 AWAWENGKWAWK-NH2 AWAWENGKWAWK-NH2 AWAWENGKWAWK-NH2
a Underlined (bold) residues are CdO. Turn residues are in italics.
13
CdO labeled on the amide
SCHEME 1: Single Isotope Labeled Variants of TZ2C Used for This Studya
a 13
C substituted CdO groups are indicated with boxes.
positions and a shift of the amide II band (down to ∼1450 cm-1). The lyophilized peptides were dissolved in D2O, which yields an acidic solution (measured pH, or pH* ∼ 1.1, employed in this paper, equivalent to pD ∼1.5), used to obtain the data reported here. Equilibrium Thermal Unfolding Measurements. Equilibrium behavior of folding and unfolding of these peptides as a function of temperature was previously studied with FTIR and CD spectroscopy at neutral pH and reported recently.8 The acidic conditions used in this study yield more reversible thermal refolding for the Trpzip2C peptides at the concentrations used for the T-jump measurements, which is required since several thousand single absorption transients have to be averaged to obtain a good signal-to-noise ratio at the several temperatures and wavelengths studied. Equilibrium FTIR data were remeasured at low pH to ensure sample continuity and to have the same pH conditions as for the T-jump experiments, since T-jump probe wavelengths have been deduced from them. Comparison of absorption spectra at neutral and acidic pH shows only minor amide I′ frequency shifts, < (2 cm-1. Samples for IR measurements at pH* ∼ 1.1 were prepared at ∼20 mg/mL in D2O. The solutions were held in a demountable cell with CaF2 windows separated by a 100 µm Teflon spacer, and temperature was controlled by flow from a thermal bath. IR measurements were performed by heating the sample cell from 5 to 90 °C (in steps of 2 or 5 °C) and cooling back down, under control of the
11630
J. Phys. Chem. B, Vol. 114, No. 35, 2010
spectrometer (Bruker Equinox 55). The amide I′ spectra between 1700 and 1600 cm-1 were deconvoluted into three bands (two for the low and high β-hairpin component and one for the unfolded component) for the unlabeled TZ2C and four bands (one more for the 13CdO component) for the labeled variants. A linear combination of Gaussian and Lorentzian bands with different line widths was used to approximate the Voigt profile for the fits.41,42 For each band component, the area, width, position of the frequency maximum, and Gaussian/Lorentzian ratio were allowed to vary with temperature. The equilibrium transition temperature, Tm value, was determined assuming a two-state model for a transition from a folded to unfolded state using the variation in fitted band area. Temperature Jump Experiments. T-jump experiments were realized with a modification of an instrument that has been described separately in our previous peptide T-jump reports.31,43 Probe and pump lasers have subsequently been improved in intensity and beam quality. Remeasurement of the TZ2C peptide relaxation kinetics using this improved instrument gave results with higher S/N levels and yielded relaxation rates that are consistent with our earlier reports. In summary, rapid heating of the solvent is induced by the Raman-shifted (to 1.9 µm) pulsed output of a Nd:YAG laser (Continuum SureLite III-10 with 5 ns pulse length). The pump beam was split into two components to excite the sample through the front and back surfaces, and each beam was focused beyond the sample to control power densities over an excitation cross section of ∼1 mm2.44-46 The output beam from a lead salt laser diode (Laser Components) was passed through a monochromator (Mu¨tek TLS 310) to select one mode, which provided the tunable IR probe. The transient transmission was measured with a photovoltaic 20 MHz MCT detector (Kolmar KMPV11-1-J2) and digitized with a transient recorder board (Imtec T3012). After normalization to the original transmitted intensity and correction for background radiation and detector drift signals (as measured by blocking the probe beam), this signal was converted to the differential absorbance, ∆A. The samples were equilibrated to a series of defined initial temperatures before the T-jump. After the pulsed laser heating (∆T ∼ 10 °C), the spectral response was monitored at selected single wavelengths in the amide I′ region as a function of time. The size of the T-jump is derived from the absorbance change of the D2O solvent that is induced by the laser pulse, as we have described separately,31,43 and is calibrated by comparison to temperature-dependent equilibrium FTIR measurements. The temperature within the probed volume changes less than 6% over 100 µs and slowly decays with more complex kinetics taking >100 ms to return back to its initial value, as was established in our original instrument and design tests.31,43 Our measurements were performed starting from different initial temperatures of the sample, and relaxation kinetics are reported here in terms of the final temperatures. Relaxation constants were determined by a double exponential fit to the signal between 500 ns and 180 µs, a span over which the fast component represents the sample relaxation and the slow one corresponds to the initial cooling of the heated volume of the solution (that can be approximated by an exponential thermal decay for times 200 µs), as we have described previously.31,43 Since the cooling rate of the solvent is very slow (∼ms), compared to the conformational changes, it has minimal interference with the determination of the hairpin dynamics (∼µs). Thus, our fits determine only one observed (fast) relaxation time for this molecular process; more were not resolvable.
Hauser et al. Calculational Methods. Quantum mechanical (QM) simulations of the IR spectra were performed using a model β-hairpin peptide structure taken from the TZ2C NMR structure.8 The structure consisted of the full backbone (12 residues) substituted with methyl side chains (yielding an all-Ala sequence, except Gly-7), which was then geometry optimized with constrained torsions and used to directly compute spectral frequency and intensity parameters (FF and APT) using Gaussian 03.47 These were transferred to a set of programs obtained from Petr Bour (Academy of Science, Prague) for spectral simulation and isotopic substitution.20,48,49 Computations on this full model hairpin, with no need for fragmentation, were done at the 6-31G*/BPW91 level using a Linux-based machine, typically with four 64-bit processors and shared memory (16 GB). Simulations were performed both for the peptide in vacuum and with additional computations using a continuum solvent correction model (COSMO, termed CPCM in Gaussian)50,51 to simulate general solvent effects on the spectra. These are useful for estimating frequency shifts due to perturbation of the diagonal FF by general solvent interaction (dielectric response), but they do not include H-bonding, so that they only provide an average, nondirected representation rather than explicit solvent effects.12,52,53 Some empirical efforts to correct the FF for solvent H-bonding to specific CdO groups were also carried out to see if the mode properties of the 13CdO vibration were affected. Results Equilibrium Unfolding Results. Trpzip2C hairpin peptides labeled on the 1, 3, and 10 positions are denoted as A1, A3, and A10. The 3 and 10 positions have CdO groups lying between the four Trp residues, in the center of the hairpin β-strands, and are expected to be the most stable part of those segments. Position 1 is on the N-terminus which is partially frayed and shows a distortion from a regular structure in our NMR studies.8,35,39 As shown in Figure 1, the difference IR spectra for A3 and A10 give a well-defined 13CdO peak at ∼1614 cm-1 that is not present in the unlabeled peptide (Figure 2). The single labeled hairpins give much weaker isotope sidebands in conventional IR absorption spectra (Supporting Information, Figure S1) than seen with the double labeled peptides A1A10 and A3A10.8,31 Consequently, the difference spectra representation makes the relatively weak IR band component for the isotope labeled variants easier to identify and facilitates identification of the optimal wavelength to be used for monitoring T-jump experiments. The double label variants have an enhanced intensity in the 13CdO modes due to cross-strand coupling with each other and with some of the 12 CdO modes. However, the A1 variant has no identifiable 13CdO component, and its normal amide I′ IR spectrum initially looks similar to that of the unlabeled variant, but with difference spectra we can identify a measurable (6 cm-1) shift of the 12C labeled A1 amide I′ IR down in wavenumber from that of the unlabeled hairpin (Figure 2), confirming the difference between the two peptides. This is in contrast to the double labeled variant A1A10 which yields a complex and unresolved but distinct amide I′ pattern for the 12C and 13C contributions.8,31 Thermal analyses of these low pH spectra yield somewhat different values for Tm and ∆H than were previously reported for studies at neutral pH, which is consistent with the pH dependence of the structural stability that we have found for the parent and related hairpin peptides.8,35 In general, at acidic pH, the peptide is less stable, as evidenced by lower Tm values.
Comparison of Isotopic Substitution Methods
J. Phys. Chem. B, Vol. 114, No. 35, 2010 11631
Figure 1. Temperature variation of amide I′ IR spectra of the single labeled TZ2C on the A3 (left) and A10 (right) positions (label between the Trp’s) at low pH, represented as difference spectra (A(T)-A(5 °C)). Temperature increases stepwise from low (blue) to high (red) temperatures (5-90 °C; only selected values plotted).
Figure 2. Temperature variation of amide I′ IR spectra of the single labeled TZ2C on the A1 (label at N terminus) position (left) and the unlabeled TZ2C (right), at low pH, plotted as in Figure 1.
The Tm values obtained for the single label species are also lower than those obtained with the double label species, which may be a problem of fitting such a multicomponent spectral variation with a single global function, since the smaller contribution of the 13CdO band depends on different local variations and the larger 12CdO contribution is also affected by removal of those local contributions, but their impact on various spectral parameters is not in balance.8 The actual Tm values obtained from the fits depend on the model used, which cannot be chosen unambiguously due to insufficient sigmoidal character for the thermal transition of these Trpzip2C hairpin peptides, since they do not become fully unfolded in our accessible temperature range. However, the overall trends in this study with acidic pH and its analysis of the amide I′ band area reflect those previously reported for the thermodynamic analysis at neutral pH using the amide I′ frequency maximum for the analysis, so that enumeration of the trends found may be of interest.8 In any of our analyses of the IR bands, the A3 and A10 peptides yielded higher Tm values than did the A1. Between these two, the A3 (Tm ) 344 ( 2 K) consistently gave higher Tm values than the A10 (Tm ) 341 ( 2 K), but they are within a standard deviation. The unlabeled TZ2C (Tm ) 337 ( 5 K) consistently had a Tm value lower than A3 and higher than A1 within the fitting error. The Tm values given above (in parentheses) refer to the analysis of the amide I′ band at acidic pH. Since the peptides are all the same sequence, these small deviations in the analyses may reflect a variation in the weighting of contributions from different residues as the labeled residue changes position. Naturally, this
would be more directly studied by focusing on the 13CdO mode. The Tm’s for the 13C amide I′ components of A3 and A10 showed differences but again within the fitting error. Theoretical Modeling of Amide I Spectrum. Using the NMR structure of Trpzip2C, we have constructed an all-Ala peptide having the same φ,ψ angles and thus the same fold and cross-strand amide coupling and H-bonds as does the original hairpin. This reduced hairpin model was then used to calculate the vibrational force field after optimizing all geometrical parameters except the dihedral angles at the DFT/BPW91/631G** level (in vacuum and then again with a global solvent correction, COSMO50,51).8 Only the COSMO results will be discussed here, and the vacuum results are provided in the Supporting Information, Figure S2 and Table S1, for comparison. These calculated results for the unlabeled peptide are in reasonable agreement with the experimental spectra (Figure 3). The experimental spectra are broader and have a more pronounced high frequency shoulder, the latter of which could possibly be enhanced by TFA contamination. The peptide in solution has a dynamic structure with heterogeneously broadened spectra that is only approximately represented with the static approach implicit in the single structure QM simulation. This difference is illustrated by the overlap spectra in Figure 3 comparing the experimental and COSMO corrected calculational results. As normal for DFT calculated FFs, the calculated amide frequencies are high, having maxima for the unlabeled variant at ∼1663 and ∼1690 cm-1, for the solvent corrected and vacuum results, respectively, while the TZ2C experimental maximum is ∼1635 cm-1 (Table 2).
11632
J. Phys. Chem. B, Vol. 114, No. 35, 2010
Hauser et al.
Figure 3. Experimental (A, top) and DFT simulated (B, bottom, with COSMO solvent correction) amide I′ absorption spectra of unlabeled TZ2C (blue) and labeled variants A1 (black), A3 (red), and A10 (green). Experimental data were measured at 5 °C, 20 mg/mL (conc. ∼12 mM) and 100µ path, and linewidths for the component bands in the simulation were chosen to mimic those of the experimental 13CdO band. (There may be minor contribution at 1672 cm-1 from residual TFA in some of the experimental spectra.) Simulations are expressed in molar absorbance, ε, or M-1 cm-1. Representation of modes is available in Supporting Information, Figure S2.
While frequency determination is important for interpreting single isotope labeled peptide spectra, the amide I′ shift of the 13 C component from the 12C main band is relatively independent of the general DFT error (compare vacuum and COSMO results in Table 2) and can distinguish site variations in the diagonal FF predicted by our model. The band shapes for both calculations are similar, implying the off-diagonal terms are not very different, but the COSMO correction leads to less dispersion of the amide I′ and more resolution of the 13C component (see Figure S2, Supporting Information). This presumably arises from a relatively larger frequency reduction for those CdO groups pointed away from the other strand (hence out into vacuum), which can alter details of individual mode mixing. To further test this concept, we empirically altered the FF for those CdO groups pointing out into solution to mimic H-bonds to the water solvent. The effect on the 13CdO mode was not detectable for
both the single and double labeled variants studied here. However, for the 12CdO modes, the mode dispersion is further reduced and the frequency of the absorbance maximum slightly shifted down ∼2 cm-1, resulting in marginally smaller 13C-12C splittings, which in principle are in better agreement with experiment but result in a more mismatched representation of the experimental bandwidth. This again suggests that conformational flexibility, or heterogeneous broadening due to an ensemble of structures, is the key missing component in simulating the observed bandwidth.14 These empirical results are summarized in the Supporting Information, Figure S3 and Table S2. Our analyses focus on 12C-13C shifts, assuming that the relative shift of 13CdO modes for different substitution positions is relatively independent of the DFT errors. This assumption has been supported by previous work on labeled peptides having both helical and hairpin conformations.1,5,7,8,12-15 Labeling a residue in the middle of a β-strand tends to disrupt the exciton coupling, which leads to less intensity redistribution into the lowest energy mode, as is typical of β-sheet structures.54 This is seen in our simulations, where there is more intensity in two modes adjacent to the strongest low frequency one causing the A3 and A10 12CdO band envelop maxima to shift up from the A1 and TZ2C maxima, but this effect is less evident (smaller shift) in the experimental data (Figure 3). In summary, increasing solvent effect corrections (from vacuum to COSMO to empirical) reduces the 12C-13C shift values, by shifting the 12CdO band down due to change in some diagonal FF elements, but does not alter their relative ordering for different substitution patterns (Table 2). The values shown in Table 2 summarize the frequencies for resultant bands corresponding to the 12C and 13C amide I′ components as computed (DFT/BPW91/6-31G**) for the TZ2C NMR structure in a vacuum and using the COSMO solvent correction, along with experimental results for single labeled hairpins. As compared to experiment, these frequencies are systematically too high by 30-40 and 40-50 cm-1 for the 13C and 12C maxima, respectively, in a vacuum and by
