Discrepancies between Conformational Distributions of a Polyalanine


Dec 7, 2010 - Maik H. Jacob , Roy N. Dsouza , Indrajit Ghosh , Amir Norouzy , Thomas Schwarzlose , and Werner M. Nau. The Journal of Physical Chemistr...
0 downloads 0 Views 1007KB Size


J. Phys. Chem. B 2010, 114, 17201–17208

17201

Discrepancies between Conformational Distributions of a Polyalanine Peptide in Solution Obtained from Molecular Dynamics Force Fields and Amide I′ Band Profiles Daniel Verbaro,† Indrajit Ghosh,‡ Werner M. Nau,‡ and Reinhard Schweitzer-Stenner*,† Department of Chemistry, Drexel UniVersity, 3141 Chestnut Street, Philadelphia, PennsylVania 19104, United States, and School of Engineering and Science, Jacobs UniVersity Bremen, Campus Ring 1, D-28759 Bremen, Germany ReceiVed: September 30, 2010; ReVised Manuscript ReceiVed: NoVember 13, 2010

Structural preferences in the unfolded state of peptides determined by molecular dynamics still contradict experimental data. A remedy in this regard has been suggested by MD simulations with an optimized Amber force field ff03* (Best, R.; Hummer, G. J. Phys. Chem. B 2009, 113, 9004-9015). The simulations yielded a statistical coil distribution for alanine which is at variance with recent experimental results. To check the validity of this distribution, we investigated the peptide H-A5W-OH, which with the exception of the additional terminal tryptophan is analogous to the peptide used to optimize the force fields ff03*. Electronic circular dichroism, vibrational circular dichroism, and infrared spectroscopy as well as J-coupling constants obtained from NMR experiments were used to derive the peptide’s conformational ensemble. Additionally, Fo¨rster resonance energy transfer between the terminal chromophores of the fluorescently labeled peptide analogue H-Dbo-A5W-OH was used to determine its average length, from which the end-to-end distance of the unlabeled peptide was estimated. Qualitatively, the experimental 3J(HN,CR), VCD, and ECD indicated a preference of alanine for polyproline II-like conformations. The experimental 3J(HN,CR) for A5W closely resembles the constants obtained for A5. In order to quantitatively relate the conformational distribution of A5 obtained with the optimized AMBER ff03* force field to experimental data, the former was used to derive a distribution function which expressed the conformational ensemble as a mixture of polyproline II, β-strand, helical, and turn conformations. This model was found to satisfactorily reproduce all experimental J-coupling constants. We employed the model to calculate the amide I′ profiles of the IR and vibrational circular dichroism spectrum of A5W, as well as the distance between the two terminal peptide carbonyls. This led to an underestimated negative VCD couplet and an overestimated distance between terminal carbonyl groups. In order to more accurately account for the experimental data, we changed the distribution parameters based on results recently obtained for the alanine-based tripeptides. The final model, which satisfactorily reproduced amide I′ profiles, J-coupling constant, and the end-to-end distance of A5W, reinforces alanine’s high structural preference for polyproline II. Our results suggest that distributions obtained from MD simulations suggesting a statistical coil-like distribution for alanine are still based on insufficiently accurate force fields. Introduction The canonical view of the unfolded state of proteins and peptides suggests that they are unstructured in that all natural amino acid residues with the exception of proline sample the entire sterically allowed region of the Ramachandran plot with comparable probability. This view is generally termed the random or statistical coil model.1-3 However, recent experiments on short peptides and analyses of truncated coil libraries (helices and sheet structures were omitted) indicated that some amino acids have a much larger preference for polyproline II (PPII) like conformations than predicted by respective Ramachandran plots obtained from molecular mechanics and dynamics calculations.4-11 The canonical PPII conformation adopted in crystals of poly-L-proline exhibits dihedral angles of about φ ) -70 and ψ ) 150.12 Among the amino acids for which preferences for PPII-like conformations have been proposed, alanine has attracted particular attention. The focus on this amino acid residue reflects its abundance in proteins, its high helical propensity in folded * Corresponding author. Phone: 1-215-895-2268. Fax: 1-215-895-1265. E-mail: [email protected] † Drexel University. ‡ Jacobs University Bremen.

structures, and the simplicity of its side chain, which facilitates computational modeling. Two-dimensional IR, analyses of the amide I band profiles of IR, Raman, and vibrational circular dichroism (VCD) spectra, NMR, and electronic circular dichroism (ECD) spectroscopy on alanine-based peptides have provided ample evidence for the notion that alanine has a preference for PPII, but attempts to quantitatively determine the propensity for this conformation yielded rather different results (i.e., mole fractions between 0.3 and 0.9 per residue).13-19 Only recently, a very thorough NMR study by Graf et al., which utilized the φ and ψ dependence of seven different dipolar J-coupling constants in a study of various oligo-alanines, resolved the issue by providing strong evidence for a very high PPII propensity of alanine (i.e., 0.9).20 This result was recently confirmed by a reanalysis of amide I profiles for A3 and by a combined NMR/ vibrational spectroscopy study on GAG.21,22 The above results for alanine are still at odds with the results of many molecular dynamics simulations on short alaninecontaining peptides which generally predict rather statisticalcoil-like distributions with a substantial fraction of right-handed helical conformation.9,10,23 This is at variance with the rather small nucleation constants generally observed from investigations of helix T coil transitions.22 The respective distribution

10.1021/jp109404r  2010 American Chemical Society Published on Web 12/07/2010

17202

J. Phys. Chem. B, Vol. 114, No. 51, 2010

depends heavily on the choice of the force field.9,10 Only a modified AMBER force field, for which Garcia and co-workers eliminated the torsional energy terms, predicts a rather high PPII and a comparatively low helical fraction.24,25 Another attempt to reconcile experimental data with MD simulations has recently been undertaken by Best and Hummer.23 They used two modifications of the original AMBER force field termed ff99B and ff03. They were both obtained from reparametrizing the AMBER force field based on results from quantum chemical calculations. Best and Hummer optimized these force fields further by trying to account for the helix content of short peptides. They subsequently checked the validity and applicability of their new force fields (termed ff99SB* and ff03*) by calculating the J-coupling constants, which Graf et al. obtained for pentaalanine in water.20 This analysis yielded a much lower PPII content than that reported by Graf et al. (0.5 for ff03* and 0.4 for ff99SB*). Generally, the distributions emerging from the calculations with both force fields resemble a classical statistical coil. Besides using different modeling techniques, Graf et al. and Best and Hummer used different coefficients for the Karplus equations employed to simulate the different J-coupling constants. Whereas Graf et al. used empirically determined coefficients, Best and Hummer obtained their values from DFT calculations on short alanine peptides (AcA and AA).26 In view of the discrepancy between the results obtained from these two sets of Karplus equations, a further experimental check of the respective distributions seems to be a necessity. To this end, we measured the amide I′ band profile of the IR and VCD spectrum of H-A5W-OH which closely resembles the oligoalanine peptide investigated by Graf et al.20 1H NMR measurements were performed to determine the 3J(HN,HR) constants of its residues. The amide I′ band profile and the NMR coupling constants reported by Graf et al. were calculated in terms of a statistical ensemble built on the results of the MD simulations of Best and Hummer.23 In order to calculate the J-coupling constants, we followed these authors by utilizing the abovementioned DFT-based Karplus parameters reported by Case et al.26 The results of these calculations were compared with a direct conformational analysis of the above oligo-alanine based on a global fit of a conformational distribution model to amide I′ band profiles and J-coupling constants. The additional tryptophan residue at the end of the peptide serves two purposes. First, it allowed us to exactly determine the concentration of the peptide, which is important for obtaining the amide I profiles in absolute units. Second, it facilitated the comparison with another experiment designed to explore the peptide’s conformational ensemble. Recently, Nau and co-workers introduced Fo¨rster resonance energy transfer (FRET) as an ideal tool for determining the average length of short peptides by using tryptophan as donor and Dbo (2,3-diazabicyclo[2.2.2]oct-2-ene) as acceptor. This pair has a very short Fo¨rster radius (R0 ) 9 Å), which allows FRET measurements of short distances in the 10 Å domain.27,28 We used this technique to determine the average distance between the fluorophores of the labeled peptide Dbo-A5W (see Scheme 1), which can be regarded as a measure of the average end-to-end distance. We compared this value with estimations derived from the different conformational models used to account for the amide I′ profiles and NMR coupling constants. Eventually, this led to a reliable, experimentally based conformational model, which considers a high PPII propensity for alanine residues, thus supporting the results of Graf et al.20 Our results suggest that further modifications of

Verbaro et al. SCHEME 1: Molecular Structures of Investigated Acceptor Chromophore, Labeled Amino Acid, and Pentaalanine Peptide

force fields are necessary for an accurate description of the conformational manifold of even very simple and short peptides. Methods and Materials Unblocked L-alanyl-L-alanyl-L-alanyl-L-alanyl-L-alanyl-Ltryptophan (H-AAAAAW-OH, A5W) was custom synthesized by Celtek Peptides with >99.3% purity and further purified through lyophilization and dialysis using a MW500 dialysis bag to remove residual TFA (trifluoroacetic acid). With the exception of the C-terminal tryptophan, which we used for an exact determination of the peptide concentration and Fo¨rster resonance energy transfer (FRET) measurements, the peptide is analogous to the unblocked pentaalanine peptide investigated by Graf et al.20 The J-coupling constants obtained for this peptide were used by Best and Hummer to check the validity of the conformational ensembles obtained from their AMBER ff03* and ff99SB* force fields. For IR and VCD, the peptide was dissolved in D2O with a concentration of 20 mM. This is the highest achievable possible concentration in aqueous solution, which is sufficient for IR and VCD.29,30 A pD of 6.9 was obtained using the Glasoe and Long31 method with an Accumet micro size standard glass combination electrode with Ag/AgCl and an Accumet pH meter (Fisher Scientific). Vibrational spectra were taken on a BioTools Chiral IR with a 20 µm cell and an 8 cm-1 resolution. The combined spectra were collected over 720 min (VCD 680 min and IR 72 min). The temperature was kept at 25 °C using a Biotools water-cooling temperature controller. For ECD, a peptide concentration of 5 mM was prepared in D2O, which was used for direct comparison to vibrational data since conformational propensities have been shown to change if D2O is replaced by H2O.32 Spectra were taken every 5 °C between the wavelengths 180 and 280 nm on a J-810 spectropolarimeter (Jasco, Easton, MD). A 0.05 mm cell was used with 0.5 nm resolution and a scan speed of 500 nm/min. An average of 10 scans was taken at each temperature. Also, for the 1H and COSY experiments of A5W, the peptide was dissolved at 20 mM in 90% H2O and 10% D2O. The 3J(HN,HR)coupling constants were obtained with a Unitylnova 500 MHz NMR spectrometer at 25 °C. The FRET experiments were performed on the peptide H-Dbo-AAAAAW-OH (custom synthesized by Biosyntan in 96.1% purity), which contained an additional N-terminal Dbo (2,3-diazabicyclo[2.2.2]oct-2-ene)-labeled asparagine. Measurements were performed at ambient temperature (25 °C) in aerated H2O at pH 6.8 ((0.2), with reference to the donor-only labeled peptide, A5W. The fluorescence lifetimes of the peptides were

Conformational Distributions of Polyalanine Peptide

J. Phys. Chem. B, Vol. 114, No. 51, 2010 17203

Figure 1. Far-UV circular dichroism spectrum of 5 mM A5W in D2O. The arrows indicate the spectral changes with increasing temperature, which was changed in increments of 5 °C. The inset shows the difference spectrum ∆ε(80°C) - ∆ε(5°C).

measured by time-correlated single-photon-counting (FLS 920, Edinburgh Instruments Ltd.) using a PicoQuant pulsed LED (PLS-280, λexc ) 280 nm, λobs ) 350 nm, fwhm ca. 450 ps) for excitation of Trp. Steady-state emission spectra and intensities (λexc ) 280 nm) were recorded with a Cary Eclipse fluorometer (Varian). For the steady-state measurements, peptide concentrations (ca. 20 µM) were adjusted to an optical density of ca. 0.10 at the excitation wavelength. Results 3

N

R

The J(H ,C ) constant was measured for all residues in order to compare the alanines of A5W with A5. Correlation spectroscopy (COSY) only allowed identification of tryptophan 3 J(HN,CR), which was 7.53 ( 0.03 Hz. The alanine coupling constants were 5.49 ( 0.02, 5.71 ( 0.06, 5.39 ( 0.09, and 6.16 ( 0.06 Hz in no particular order. The coupling constants seem to be in close agreement with the values Graf et al. reported for A5.20 None of the alanines’ 3J(HN,CR) could be assigned to a distinct residue; however, all of the coupling constants are indicative of a φ value in the PPII region of the Ramachandran space. Electronic circular dichroism was utilized to qualitatively gauge the secondary structures adopted by A5W. The respective spectra measured as a function of temperature are displayed in Figure 1. At lower temperatures, the extrema at 220 and 190 nm are together diagnostic of a significant PPII fraction.22,29,35 The ∆ε values at these wavelengths are comparable with those reported for GAG and AAA,22,29,32 which strongly suggests a similar PPII content. Apparently, the influence of the terminal tryptophan on the far-UV CD signal is very barely detectable. As the temperature increases, the couplet amplitude decreases. The spectra clearly depict an isodichroic point, suggesting that the peptide predominantly samples only two minima of its Gibbs energy landscape. The difference spectrum ∆ε(80°C) - ∆ε(5°C) exhibited in the inset clearly suggests that β-strand-like conformations become more populated at high temperatures, in agreement with expectations.33 It is noteworthy that these ECD spectra, though yielding only qualitative information, are certainly at variance with the conformational distributions obtained from MD simulations of A5, in that the latter suggest at least a three-state model (PPII, β, right-handed helical). In what follows, this paper focuses on analyzing the conformational distribution, which A5W exhibits at room temperature. Figure 2 exhibits the amide I′ profiles of the FTIR and VCD spectra of A5W. The IR absorption band shows a peak at 1642

Figure 2. Amide I′ region of the infrared and vibrational circular dichroism spectra of A5W in D2O. The experimental conditions chosen for recording these spectra are described in Methods and Materials. The red lines result from a simulation using a conformational distribution reflecting the Ramachandran plot obtained from MD simulations with a ff03* force field. The black lines reflect the results of a fit with an adjustable conformational model describable as superposition of two-dimensional Gaussian distributions associated with PPII, β-strand, right-handed helical, and inverse γ-turn-like conformations. The conformational fractions are listed in Table 3 as amide I′based I. The blue line was computed with a refined model, which additionally considered a further modified distribution as mentioned in the results.

cm-1 and a shoulder at 1660 cm-1. The corresponding VCD spectrum depicts a strong negative couplet with a positive extremum around 1663 cm-1 and a negative one at 1638 cm-1, which is indicative of a strong preference for a PPII-like structure.34 In the following, these two profiles are used to check the validity of the conformational distributions, which Best and Hummer obtained from their MD simulations with ff03* and ff99SB* force fields.23 To this end, we proceeded as follows. We used the Ramachandran plots obtained from ff03* simulation (kindly provided by Dr. Best) to construct the following distribution function. The upper left-handed quadrant was divided into two subspaces centered at (variable) coordinates assignable to PPII and β-strand conformations. Points between (-90 < φ < -40 and 110 < ψ < 180) were considered PPII, (-180 < φ < -100 and 110 < ψ < 180) were considered β, and (-160 < φ < -20 and -120 < ψ <50) were considered right-handed helical. Conformations associated with each subspace were described by five representative sets of coordinates, which correspond to the maximum of a twodimensional Gaussian function and the corresponding half-width position along φ and ψ. The total ensemble was then described as a superposition of weighted subspace distributions. The corresponding weighting factors were taken from Best and Hummer.23 Alternatively, in order to account for the very

17204

J. Phys. Chem. B, Vol. 114, No. 51, 2010

Verbaro et al.

TABLE 1: List of Experimental and Theoretically Calculated J-Coupling Constants (Hz) of Individual Amino Acid Residues of A5 and A5Wa A5 A2 3 J(HN,HR) 3 J(HN,C′) 3 J(HR,C′) 3 J(HN,Cβ) 1 J(N,CR) A3 3 J(HN,HR) 3 J(HN,C′) 3 J(HR,C′) 3 J(HN,Cβ) 1 J(N,CR) A4 3 J(HN,HR) 3 J(HN,C′) 3 J(HR,C′) 3 J(HN,Cβ) 1 J(N,CR) A5 3 J(HN,HR) 3 J(HN,C′) 3 J(HR,C′) 3 J(HN,Cβ) 1 J(N,CR)

A5 W

exptlb

constraint MDc

ff03*d

ff03* modelede

amide I′ based If

amide I′ based Ig

amide I′ based IIh

5.59 ( 0.03 1.13 ( 0.00 1.85 ( 0.05 2.30 ( 0.00 11.36 ( 0.03

5.5 1.2 1.5 2.1 10.9

6.71 1.8 2.49 11.45

6.71 0.95 1.82 2.50 11.19

5.48 1.25 1.46 2.27 11.18

5.86 1.06 1.41 3.57 11.18

5.72 1.09 1.77 2.26 11.03

5.74 ( 0.02 1.86 ( 0.05 2.24 ( 0.01 11.26 ( 0.03

5.7 1.6 2.0 10.7

6.91 1.86 2.54 11.27

6.91 1.86 2.52 11.35

5.76 1.55 2.17 11.17

6.18 1.53 3.38 11.17

5.77 1.83 2.26 11.02

5.98 ( 0.02 1.15 ( 0.02 1.89 ( 0.02 2.14 ( 0.00 11.25 ( 0.02

5.9 1.1 1.7 2.0 10.6

7.07 0.81 1.94 2.47 11.18

7.06 0.84 1.92 2.44 11.43

6.13 1.37 1.65 2.03 11.21

6.60 1.35 1.69 3.14 11.21

5.96 1.01 1.87 2.24 11.20

6.54 ( 0.05 1.16 ( 0.06 2.19 ( 0.01 1.96 ( 0.00 11.49 ( 0.03

-

6.74 1.09 1.87 2.33 -

6.73 1.05 1.84 2.38 -

6.39 1.40 1.73 1.89 11.20

6.89 1.45 1.80 2.98 11.20

6.46 1.13 1.88 2.05 11.28

a The experimental 3J(HN,HR) for A5W are mentioned in the results as each alanine could not be distinguished. The calculated constants were derived from conformational distributions of alanine residues obtained from MD simulations with a modified AMBER force field (ff03*) and from fits to the amide I′ profiles of the IR and VCD spectrum of A5W. b Obtained from Graf et al.20 c These coupling constants emerged from a MD simulation constrained by the listed coupling constants, as described by Graf et al.20 d Obtained from a MD simulation with an ff03* force field and DFT based Karplus parameters, as described by Best and Hummer.23 e Obtained from a distribution model based on the MD simulation with an ff03* force field and DFT based Karplus parameters, as described in the text of this paper. f Obtained from a distribution model derived from an analysis of the amide I′ band profiles and calculated with empirical Karplus constants,20 as described in the text of the paper. g Obtained from a distribution model derived from an analysis of the amide I′ band profiles and calculated with DFT-based Karplus constants,26 as described in the text of the paper. h Obtained from the second modified distribution with less beta and added turn structures as listed in Table 3.

unstructured distributions obtained by these authors (as seen in Figure S3 in the Supporting Information), we substituted the Gaussian functions by two-dimensional plateau functions written as fj(θ) ) χj ·

{

1 for [φj - ∆φj e φ e φj + ∆φj ∧ ψj - ∆ψj e ψ e ψj + ∆ψj] 0 otherwise

(1)

We used these distribution functions to calculate the J-coupling constants reported by Graf et al. by means of both the DFT26 and the empirically based Karplus relations.20 Additionally, we used these distributions to calculate the amide I′ profiles in Figure 2 by means of an excitonic coupling model which describes both profiles in terms of delocalized vibrational states produced by nearest-neighbor through-bond and non-nearestneighbor through-bond coupling. The respective formalism has been described in detail in recent publications,21,22,30,34 in which we revealed that particularly the VCD signal is very sensitive to even small population changes of conformational states. For the purpose of demonstration, a comparison of the amide I′ VCD for different conformational ensembles is shown in the Supporting Information (Figure S1). All calculations were carried out using the MATLAB software. In a first step, we used the above-mentioned distributions to reproduce the J-coupling constants which Best and Hummer

derived from the conformational distribution obtained with their ff03* force field.23 This was achieved by using the central coordinates, widths, and statistical weights of the distribution functions as free parameters. Thus, we found that Gaussian and plateau type distributions with the same center, width, and fraction produced practically identical amide I profiles and J-coupling constants (Figure S2 and Table S1 in the Supporting Information). In order to achieve consistency, the DFT-obtained Karplus parameters26 utilized by Best and Hummer were employed. The final values for all J-coupling constants calculated by this procedure were found to be in good agreement with the respective values obtained from MD simulations, as shown in Table 1. This demonstrates that our distribution function represents the conformational ensembles, which these authors obtained with their modified force fields ff03*, with sufficient accuracy. Since the differences between the 3J-values and amide I′ profiles obtained with the two distribution models are insignificant, all further simulations solely used the step function. In a second step, we used the thus-obtained distribution functions to calculate the IR and VCD amide I′ band profiles. Both IR and VCD profiles depend on the dihedral angles φ and ψ of the nonterminal residues owing to the conformational dependence of nearest and non-nearest excitonic coupling and their direct dependence on the relative orientation of the transition dipole moments.34 For the intrinsic amide I′ wavenumbers of the nonterminal residues, we used 1652.5 cm-1, which is close to the value found for the central residue of

Conformational Distributions of Polyalanine Peptide tetraalanine.35 The wavenumbers of the terminal amide I′ modes were down (C-terminal) and upshifted (N-terminal) to account for obtained end effects.35 The respective transition dipole moments were also taken from Measey et al.36 Nearest-neighbor coupling parameters were obtained as described by SchweitzerStenner.21 Transition dipole coupling was used to account for non-nearest-neighbor coupling.37 The IR and VCD profiles were eventually calculated as superposition of Gaussian bands associated with individual excitonic transitions. The common half-bandwidths of all these bands were assumed to be 12 cm-1.35,38 Figure 2 displays the results of this simulation obtained with the ff03* conformational distribution. The simulated infrared band accounts for the experimentally obtained shoulder at 1660 cm-1 but does not fill the intensity around 1642 cm-1. The simulated VCD does exhibit the negative couplet, but it is significantly weaker than the experimental signal, which suggests an underestimation of PPII. It should be further noticed in this context that the rather significant overestimations of the 3 J(HN,HR) constants of the residues A2-A4 (c.f. the experimental values of 5.59, 5.74, and 5.98 Hz with the calculated values of 6.71, 6.91, and 7.07 Hz) point into the same direction. The conformational distribution utilized for the above simulation of our amide I profiles can be used to calculate the expectation value for the distance between the two terminal CO groups, as described by Schweitzer-Stenner and Measey.18 This yielded a value of 12.7 Å. In order to assess the validity of this value, we performed a FRET experiment with an A5W peptide, to which the asparagine-labeled acceptor fluorophore (Dbo) had been attached at the N terminus of the peptide. The structures of the acceptor fluorophore (DBO), its asparagine derivative (Dbo), and the doubly labeled peptide (Dbo-A5W) are shown in Scheme 1. For the calculation of the Fo¨rster radius R0, the fluorescence quantum yield of the Trp-labeled reference peptides was determined by relative measurements, using N-acetyltryptophanamide (NATA) as reference (quantum yield )0.14, in water). The quantum yield for the C-terminal W-peptide was found to be 0.085, slightly larger than for N-terminal W-peptides (0.059).27 This results in a slightly larger R0 value of 9.6 Å for the C-terminal Trp/N-terminal Dbo peptide than the R0 value of 9.0 Å previously reported for N-terminal Trp/C-terminal Dbo peptides.27,28 The local refractive index of the peptide-bound fluorophores was taken as 1.340 in water and the orientation factor was assumed to be 2/3. The very small R0 value is a special characteristic of the Trp/ Dbo FRET pair, which makes distance determinations in short peptides feasible. In addition, the absence of Dbo absorbance at 280 nm (Figure 3, extinction coefficient at 280 nm <10 M-1 cm-1) ensures a perfectly selective excitation of the donor Trp, which also greatly facilitates the experimental design and quantitative analysis of the FRET data. The relative fluorescence intensities and lifetimes of the donor/acceptor doubly labeled versus the donor-only labeled peptide are shown in Figure 4. As can be seen, the fluorescence intensity and lifetime of Trp decreases significantly in the presence of the Dbo acceptor, which can be assigned to FRET. The steady-state and time-resolved fluorescence quenching can be further used to quantify the energy-transfer efficiencies, from which effective distances between donor and acceptor can be extracted according to our previously established and detailed methodology (cf. Supporting Information).27,28 The effective donor-acceptor distance (Reff) obtained from steady-state FRET was 10.0 Å and that from time-resolved FRET was 10.1 Å (see Table 2 and Supporting Information). The above values have to be corrected for diffusion by adding ca. 1 Å. This is the value

J. Phys. Chem. B, Vol. 114, No. 51, 2010 17205

Figure 3. Spectral overlap of the C-terminal Trp emission (dotted line) with the Dbo absorption spectrum (solid line) in water, obtained from the respective single-labeled reference peptides. The Fo¨rster radius for this pair is 9.6 Å.

Figure 4. (a) Steady-state fluorescence spectra and (b) time-resolved fluorescence decays (λexc ) 280 nm, λobs ) 350 nm) of the Dbo-A5W peptide in water at pH 6.8, with reference to the donor-only substituted reference peptide, A5W. The instrument response function (IRF) for the time-resolved measurements is also shown. The average fluorescence lifetimes obtained from biexponential fitting are 1.48 ns for DboA5W and 2.24 ns for the A5W reference peptide.

determined for flexible peptides by conducting measurements in propylene glycol, where diffusion-enhanced FRET is presumed to be negligible.28 We consider the 1 Å correction as generally valid for peptides of the size of Dbo-A5W and conclude therefore that the donor-acceptor distance between Dbo and W is 11 Å. In order to relate this number to the average distance between terminal CO bonds obtained from conformational distributions, we used the molecular mechanics component of the TITAN software (Schro¨dinger, Inc.) to optimize the structure of DboA5W for a backbone conformation in which all alanines adopt either a PPII (φ ) -70°, ψ ) 150°) or a β-strand conformation (φ ) -120°, ψ ) 115°), whereas a β-strand conformation was

17206

J. Phys. Chem. B, Vol. 114, No. 51, 2010

TABLE 2: Energy-Transfer Efficiencies (E) and Calculated Effective Trp-Dbo Distances (Reff) in the Peptide Dbo-A5W Obtained from Steady-State and Time-Resolved Fluorescence Measurements in H2O at pH 6.8a measurement

E

Reff/Å

steady state time resolved

0.44 ( 0.02b 0.34 ( 0.01c

10.0 ( 0.2 10.1 ( 0.2

a See Supporting Information for details of data analysis and calculations. b Quantum yield of donor-only labeled peptide, A5W, was determined as 0.085. c Absolute fluorescence lifetimes are 1.48 ns for Dbo-A5W and 2.24 ns for the A5W reference peptide.

assumed for the terminal tryptophan. The latter is consistent with the very high 3J(HN,HR) value of the C-terminal amide proton (7.53 ( 0.03). Thus, we found that the distance between Dbo and W is between 0.5 and 1 Å larger than that between terminal peptide carbonyl groups. Hence, the CO-CO distance of 12.7 Å obtained from our modeling of the conformational distribution obtained with the ff03* force field would correspond to Dbo-W distance between 13.2 and 13.7 Å, which, when compared with the 11 Å obtained from our FRET data, must be judged as an overestimation. In order to achieve better agreement between simulated and experimental amide I′ profiles, a higher population of PPII was considered to account for the rather large negative couplet. We found that the reproduction of the latter requires an average PPII fraction of 0.75. In order to obtain a satisfactory fit for both the VCD signal and the IR band profile, the half-widths σφ and σψ of the two-dimensional Gaussian distributions had to be reduced from 20° to 10° for all three considered conformations. Moreover, the region with helical conformations was confined to angles generally found within R-helices as opposed to the large A+ region obtained by Best and Hummer.23 Additionally, we followed Hagarman et al.22 by considering the population of states with inverse γ-turn-like conformations (-80° < φ < -70° and 55° < ψ < 60°). The black solid line in Figure 2 shows the best fit to our experimental data. The rather sharp individual Gaussian distributions used for these fits have also recently been obtained for GXG, A3 and V3 peptides.21,22 Thus, our results suggest that the conformational space sampled by alanine residues is much more confined than suggested by the statistical coillike distributions obtained by MD simulations with ff03* and many other force fields.9,10,39,40 The respective fractions of PPII, β-strand, helical, and γ-turn-like conformations are listed in Table 3. Apparently, the PPII fraction decays somewhat toward the C-terminal of the peptide. While the above simulation fully accounts for our amide I′ band profiles, the respective calculated J-coupling constants, which were still obtained with DFT-based Karplus constants, do not reproduce all experimental values. This notion particularly holds for 3J(HN,Cβ). However, if the empirical parameters reported by Graf et al.20 are used, the coupling values closely reproduce the experimental data without any further refinement, as one can infer from Table 1, which lists all the distribution parameters used for this simulation. The respective distance between terminal CO groups is 11.9 Å, which would correspond to a Dbo-W distance between 12.4 and 12.9 Å. This is closer to the experimental distance measured by FRET (11 Å) than the value derived for the ff03* distribution, but the difference is still notable. This discrepancy prompted us to try a further refinement for which we shifted small fractions from the β-strand ensemble to helical and inverse γ-turn-like conformations. Additionally, we slightly shifted the position of the PPII distribution. The

Verbaro et al. TABLE 3: Fraction of Conformations Sampled by the Indicated Amino Acid Residues of A5W As Derived from the Amide I′ Profilesa PPII

β

R

γ-turn

center coord (deg) A2 A3 A4 A5 W6

Amide -70, 145 0.835 0.775 0.725 0.675 0.500

I′ Based I -120, 130 0.040 0.100 0.175 0.225 0.500

-60, -60 0.075 0.070 0.050 0.030 0.000

-70, 55 0.050 0.055 0.050 0.070 0.000

center coord (deg) A2 A3 A4 A5 W6

Amide -74, 152 0.650 0.650 0.750 0.750 0.700

I′ Based II -115, 120 0.030 0.030 0.030 0.140 0.300

-60, -30 0.220 0.200 0.100 0.050 0.000

-80, 60 0.050 0.060 0.060 0.030 0.000

a Details are explained in the main text of the paper. The half-bandwidths of the considered distributions along both dihedral coordinates are 10° except for the turn, for which only a single coordinate was considered. The fractions for W6 were obtained by invoking two coexisting representative conformations as described by Shi et al.15

best fit, which reflects a high preference for PPII-like conformations, is visualized by the blue lines in Figure 2. The agreement between fit and experiment is still satisfactory, particularly for the VCD couplet. Table 1 compares the J-coupling constants, which emerged from this refinement with the respective experimental values (denoted as amide I′-based II). Overall, the agreement between calculated and experimental J-coupling constants has been improved further by this step. This particularly holds for 3J(HR,C′) and to some extent also for 3J(HN,C′) and 3J(HN,Cβ). The average CNO-CCO distance obtained from the corresponding conformational distribution is 10.5 Å, which would correspond to average Dbo-W distances between 11 and 11.5 Å. Thus, our refined fitting yielded now an excellent agreement with all experimental data. The fractions of the considered conformations obtained from this fit are also listed in Table 3. Discussion This paper was aimed at exploring whether the conformational distribution of alanine in an unfolded oligo-alanine peptide recently obtained from MD simulations carried out with a modified AMBER force field denoted ff03* can reproduce the amide I′ band profiles and the experimentally determined endto-end distances of a polyalanine. This investigation was particularly motivated by the fact that J-coupling constants derived from this distribution reproduced experimental values, if DFT-based rather than empirical Karplus constants were utilized. The distribution resembles to a major extent the classical statistical coil picture, which Ramachandran, Flory, and associates obtained for the alanine dipeptide.1,2 It is therefore at variance with the recently advanced notion that alanine has a strong preference for PPII-like conformations.16,18,20,21 Our analysis showed that the ff03*-based ensemble of conformations cannot reproduce the VCD and IR band profiles of the amide I′ mode of the hexamer A5W and that it overestimates the interfluorophore distance between Dbo and W in the related Dbo-A5W peptide. Instead, a conformational distribution which resembles to a major extent the recently reported PPII-dominated ensembles was shown to reproduce the amide I′ profiles and also the J-coupling constants, provided that the latter were

Conformational Distributions of Polyalanine Peptide calculated with empirical rather than with DFT-based Karplus constants. Moreover, this distribution leads to estimates of the average Dbo-W distance adopted by Dbo-A5W, which is closer to the experiment than the value derived from the ff03* distribution. We think that our results suggest that further modifications of MD force fields are necessary to account for all experimental data available for alanine and other amino acid residues. Thus far, only the more radically changed AMBER force field utilized by Gnanakaran and Garcia24sthe authors eliminated the torsional potentials for φ and ψsled to predictions of PPII propensities close to the experimentally obtained values. MD simulations with many other force fields predict either a statistical coil or even a dominance of helical conformations. A crucial issue with regard to the comparison of MD-derived distributions and J-coupling constants derived from NMR experiments is the use of valid Karplus parameters. Over a long period of time NMR spectroscopists have relied on empirically determined parameters for 3J(HN,HR), which have been constantly refined. In recent years, they were augmented by empirical Karplus parameters for other J-coupling constants,20 which have been used for exploring the structural ensembles of unfolded peptides and proteins. An alternative methodology has recently been developed which relies on DFT- rather than empirically based Karplus constants. While an earlier analysis of amide I profiles and J-coupling constants of A3 yielded very similar results with both sets of Karplus parameters,21 the current study argues more in favor of the empirical values. MD simulations of short unfolded peptides have thus far focused mostly on alanine.9,10,24,25,41,42 Since this residue seems to be a special case with regard to its very restricted sampling of the Ramachandran space, we are wondering whether MD simulations for other residues might yield results closer to experimentally determined propensities. In this regard, a recent paper of Beck et al. is noteworthy.40 These authors used a novel program termed ENCAD to perform MD simulations aimed at exploring the conformational distributions of all 20 amino acid residues in a GGXGG host-guest system. Their results suggest statistical coil distributions for all residues. Unfortunately, none of their propensity values agree with recently reported experimental data.15,20-22,43 The reported PPII fraction for alanine lies below 0.2 which contradicts the values of practically all experimental studies reported thus far. We used their population frequency values to calculate the IR and VCD amide I′ band profiles for A5W. The fraction in the upper left quadrant containing PPII and β conformations is only about 0.335. We used different PPII/β-strand ratios consistent with this number to calculate the amide I′ IR and VCD profiles for A5W. The results of this calculation are plotted in Figure 5. They clearly show that this conformational distribution does not come even close to representing our experimental amide I′ profiles. Since the right-handed R-helical fraction emerging from these MD simulations is large, all simulations shift the first moment of the IR amide I band to higher wavenumbers compared with results obtained with the ff03* and our own distribution. Concomitantly, all calculations yielded a positive couplet for the VCD of amide I. As seen in Table 4, the J-coupling values obtained for a 50:50 mixture of beta and PPII are not too different from experimental values for GAG. These results demonstrate again that different J-coupling constants alone, while indispensable, do not yield a unique picture of conformational distributions for the investigated residues. Augmenting them by other data like amide I profiles,

J. Phys. Chem. B, Vol. 114, No. 51, 2010 17207

Figure 5. Simulations of the amide I′ IR and VCD band profiles of A5W based on the conformational distributions of alanine reported in Beck et al.23 Simulations have been performed for different PPII/β strand ratios for the fraction assignable to the upper left quadrant of the Ramachandran plot. The following conformational ensembles were used. Black line: 0.335 PPII, 0.585 R right-handed, and 0.08 R lefthanded. Red line: 0.335β-strand, 0.585R-helical right-handed, and 0.08R-helical left-handed. Blue line: 0.1675 PPII, 0.1675 β-strand, 0.585 R-helical right-handed, and 0.08 R-helical left-handed.

TABLE 4: Coupling Constants Obtained Using the Beck et al. Distribution Compared to Experimental Data J(HN,HR) J(HN,C′) 3 J(HR,C′) 3 J(HN,Cβ) 1 J(N,CR) 3

3

100ppIIa

100βa

50ppII:50βa

exptl GAG (Hz)b

4.74 1.31 1.70 2.33 10.12

6.31 1.62 2.17 1.71 9.77

5.52 1.47 1.94 2.02 9.95

6.11 ( 0.02 1.18 ( 0.07 2.02 ( 0.10 2.32 ( 0.06 11.28 ( 0.07

a Obtained by using the average distribution for alanine from Beck et al.23 and calculated empirical Karplus parameters.20 b Experimental data from Hagarman et al.22

end-to-end distances, and, for a qualitative assessment, also UVCD spectra remains a necessity.44 Taken together, our simulations reinforce the notion that the dominant conformation of alanine in polyalanine peptides is PPII. MD simulations suggesting more statistical coil-like distributions cannot be reconciled with spectroscopic data. This underscores that a broad set of data from different spectroscopic techniques must be used to obtain a reliable picture of the unfolded state of peptides and thus also of proteins. Acknowledgment. This work was supported by a grant from the National Science Foundation (NSF, Chem 0804492, and an REU supplement, Chem 0939972) to R.S.S.. We also thank Dr. Best for providing the conformational distributions obtained by the optimized force field ff03*. I.G. and W.M.N. thank the

17208

J. Phys. Chem. B, Vol. 114, No. 51, 2010

Deutsche Forschungsgemeinschaft (DFG, NA 686/6) for support of this collaborative project. Supporting Information Available: Calculated VCD and IR spectra as a function of various conformational mixtures and details on the analysis of the FRET data. This material is available free of charge via the Internet at http://pubs.acs.org. References and Notes (1) Ramachandran, G. N.; Ramachandran, C.; Sasisekharan, V. J. Mol. Biol. 1963, 7, 95. (2) Brant, D. A.; Flory, P. J. J. J. Am. Chem. Soc. 1965, 87, 2791. (3) Flory, P. J. Statistical Mechanics of Chain Molecules; Wiley & Sons: New York, 1969. (4) Serrano, L. J. Mol. Biol. 1995, 254, 322. (5) Jha, A. K.; Colubri, A.; Zaman, M. H.; Koide, S.; Sosnick, T. R.; Freed, K. F. Biochemistry 2005, 44, 9691. (6) Jha, A. K.; Kolubri, A.; Freed, K. F.; Sosnick, T. R. Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 13099. (7) Avbelj, F.; Baldwin, R. L. Proc. Natl. Acad. Sci. U.S.A. 2003, 100, 5742. (8) Swindells, M. B.; MacArthur, M. W.; Thornton, J. M. Nat. Struct. Biol. 1995, 2, 596. (9) Duan, Y.; Wu, C.; Chowdury, S.; Lee, M. C.; Xiong, G.; Zhang, W.; Yang, R.; Cieplak, P.; Luo, R.; Lee, T.; Caldwell, J.; Wang, J.; Kollman, P. J. Comput. Chem. 2003, 24, 1999. (10) Zagrovic, B.; Lipfert, J.; Sorin, E. J.; Millett, I. S.; van Gunsteren, W. F.; Doniach, S.; Pande, V. S. Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 11698. (11) Hovmo¨ller, S.; Zhou, T.; Ohlson, T. Acta Crystallogr. 2002, D58, 768. (12) Cowan, P. M.; McGavin, S. Nature 1955, 176, 501. (13) Woutersen, S.; Hamm, P. J. Phys. Chem. B 2000, 104, 11316. (14) Woutersen, S.; Hamm, P. J. Chem. Phys. 2001, 114, 2727. (15) Shi, Z.; Chen, K.; Liu, Z.; Ng, A.; Bracken, W. C.; Kallenbach, N. R. Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 17964. (16) Shi, Z.; Olson, C. A.; Rose, G. D.; Baldwin, R. L.; Kallenbach, N. R. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 9190. (17) Eker, F.; Cao, X.; Nafie, L.; Schweitzer-Stenner, R. J. Am. Chem. Soc. 2002, 124, 14330. (18) Schweitzer-Stenner, R.; Measey, T. Proc. Natl. Acad. Sci. U.S.A. 2007, 104, 6649.

Verbaro et al. (19) Makowska, J.; Rodziewicz-Motowidlo, S.; Baginska, K.; Vila, J. A.; Liwo, A.; Chmurzynski, L.; Scheraga, H. A. Proc. Natl. Acad. Sci. U.S.A. 2006, 103, 1744. (20) Graf, J.; Nguyen, P. H.; Stock, G.; Schwalbe, H. J. Am. Chem. Soc. 2007, 129, 1179. (21) Schweitzer-Stenner, R. J. Phys. Chem. B 2009, 113, 2922. (22) Hagarman, A.; Measey, T. J.; Mathieu, D.; Schwalbe, H.; Schweitzer-Stenner, R. J. Am. Chem. Soc. 2010, 132, 540. (23) Best, R. B.; Hummer, G. J. Phys. Chem. B 2009, 113, 9004. (24) Gnanakaran, S.; Garcia, A. E. J. Phys. Chem. B 2003, 107, 12555. (25) Garcia, A. E. Polymer 2004, 120, 885. (26) Case, D. A.; Scheurer, C.; Bru¨schweiler, R. J. Am. Chem. Soc. 2000, 122, 10390. (27) Sahoo, H.; Roccatano, D.; Hennig, A.; Nau, W. M. J. Am. Chem. Soc. USA 2007, 129, 9762. (28) Sahoo, H.; Roccatano, D.; Zacharias, M.; Nau, W. D. J. Am. Chem. Soc. 2006, 128, 8118. (29) Schweitzer-Stenner, R.; Measey, T. J.; Hagarman, A.; Dragomir, I. The Structure of Unfolded Peptides and Proteins Explored by Vibrational Spectroscopy In Instrumental Analysis of Disordered Proteins: assessing Structure and Conformation; Uversky, V., Longhi, S., Eds.; Wiley & Sons, Inc: New York, 2010; pp 171-226. (30) Schweitzer-Stenner, R. Vibr. Spectrosc. 2006, 42, 98. (31) Glasoe, P. K.; Long, F. A. J. Phys. Chem. 1960, 64, 188. (32) Eker, F.; Griebenow, K.; Schweitzer-Stenner, R. J. Am. Chem. Soc. 2003, 125, 8178. (33) Yang, W. Y.; Larios, E.; Gruebele, M. J. Am. Chem. Soc. 2003, 125, 16220. (34) Schweitzer-Stenner, R. J. Phys. Chem. B 2004, 108, 16965. (35) Schweitzer-Stenner, R.; Eker, F.; Griebenow, K.; Cao, X.; Nafie, L. J. Am. Chem. Soc. 2004, 126, 2768. (36) Measey, T.; Hagarman, A.; Eker, F.; Griebenow, K.; SchweitzerStenner, R. J. Phys. Chem. B 2005, 109, 8195. (37) Torii, H.; Tasumi, M. J. Raman Spectrosc. 1998, 29, 81. (38) Schweitzer-Stenner, R. Biophys. J. 2002, 83, 523. (39) Best, R. B.; Buchete, N. V.; Hummer, G. Biophys. J. 2008, 95, L07. (40) Beck, D. A. C.; Alonso, D. O. V.; Inoyama, D.; Dagget, V. Proc. Natl. Acad. Sci. U.S.A. 2008, 105, 12259. (41) Mu, Y.; Stock, G. J. Phys. Chem. B 2002, 106, 5294. (42) Mu, Y.; Kosov, D. S.; Stock, G. J. Phys. Chem. B 2003, 107, 5064. (43) Shi, Z.; Shen, K.; Liu, Z.; Kallenbach, N. R. Chem. ReV. 2006, 106, 1877. (44) Woody, R. W. J. Am. Chem. Soc. 2009, 131, 8234.

JP109404R