Conformation of Polyalanine and Polyglycine Dications in the Gas

Jun 10, 2010 - The conformation of model [Arg(Ala)4X(Ala)4Lys+2H]2+ and [Arg(Gly)4X(Gly)4Lys+2H]2+ peptides has been systematically investigated as a ...
1 downloads 0 Views 3MB Size
6888

J. Phys. Chem. A 2010, 114, 6888–6896

Conformation of Polyalanine and Polyglycine Dications in the Gas Phase: Insight from Ion Mobility Spectrometry and Replica-Exchange Molecular Dynamics Florian Albrieux,†,‡ Florent Calvo,† Fabien Chirot,*,‡ Aleksey Vorobyev,§ Yury O. Tsybin,§ Vale´ria Lepe`re,†,| Rodolphe Antoine,† Je´roˆme Lemoine,‡ and Philippe Dugourd† UniVersite´ de Lyon, F-69622, Lyon, France, UniVersite´ Lyon 1, Villeurbanne, CNRS, UMR 5579, LASIM, UMR 5180, LSA, and Biomolecular Mass Spectrometry Laboratory, Ecole Polytechnique Fe´de´rale de Lausanne, CH-1015 Lausanne, Switzerland ReceiVed: March 23, 2010; ReVised Manuscript ReceiVed: May 19, 2010

The conformation of model [Arg(Ala)4X(Ala)4Lys+2H]2+ and [Arg(Gly)4X(Gly)4Lys+2H]2+ peptides has been systematically investigated as a function of the central amino acid X through a combined experimental and theoretical approach. Mass spectrometry-based ion mobility measurements have been performed together with conformational sampling using replica-exchange molecular dynamics to probe the influence of each amino acid on the stable peptide conformation. Satisfactory agreement is obtained between measured and calculated diffusion cross section distributions. The results confirm the propensity of alanine-based peptides to form R-helices in the gas phase, differences between peptides arising from the local arrangement of the central side chain with respect to the charged ends. More generally, we find that charge solvation plays a major role in secondary structure stabilization, especially in the case of glycine-based peptides. The rich variety of conformations exhibited by the latter is qualitatively captured by the simulations. This work illustrates the potentiality of such combined experimental/theoretical strategy to determine peptide secondary structures. The present polyalanine and polyglycine peptides also offer a series of benchmark systems for future conformation-resolved studies. I. Introduction Understanding the factors that stabilize secondary-structure elements like R-helices and β-sheets is critical to understanding protein folding, which makes it relevant to conformational diseases and more generally to the drug design issue. The conformation adopted by a molecule depends on its potential energy landscape.1 For a peptide, it is a multidimensional surface influenced by many internal and external factors.2 In particular, solvent effects can be put aside by studying molecules in vacuo.3-5 Some amino acids are known to favor specific local arrangements of the peptide backbone, as reflected by values of the (φ, ψ) dihedral angles. For example, alanine has a high propensity to form helices in the gas phase6-8 and in solution,9,10 which is not the case for glycine.11-13 The interaction of amino acids with more distant neighbors and their ability to form H-bonds with other side chain residues or backbone atoms also contribute in stabilizing secondary structures. Besides the nature of the residue, the influence of charge state and location, cation complexation, hydration, and temperature on conformation have also been examined.6,8,14,15 Numerous fragmentation techniques are available to determine experimentally the primary structure of biomolecules in the gas phase, ranging from collision-induced dissociation16 to electronic excitation methods.17-21 Although hydrogen/deuterium exchange provides general information about the conformation of isolated peptides,22 only a few techniques are able to * Corresponding author. Phone: 334 724 327 80. E-mail: fabien.chirot@ univ-lyon1.fr. † Universite´ de Lyon. ‡ LSA. § Ecole Polytechnique Fe´de´rale de Lausanne. | Present address: Universite´ Paris Sud, Institut des Sciences Mole´culaires d’Orsay F-91405, Orsay, France.

determine their detailed secondary structure. Infrared spectroscopy (IR/UV or IR multiphoton dissociation, IRMPD) is especially useful for probing local conformational preferences of amino acids in short neutral or ionic peptides.23,24 Recently, the whole structure of peptides could be elucidated using such techniques owing to systematic isotopic substitution of the nitrogen atoms.25 By measuring electric dipole moments, beamdeflection experiments have also provided indirect structural information.11,15 Finally, as pioneered by the groups of Jarrold26 and Bowers,27 direct insight on molecular shape in the gas phase can be obtained using ion mobility spectrometry (IMS) through the determination of diffusion cross sections in a buffer gas. Structural assignment by any of the above techniques relies on satisfactory comparison with calculated conformations. The relevant theoretical methods in this context encompass molecular mechanics (possibly with different levels of coarse-graining) and explicit electronic structure descriptions (most often using density functional theory). While the latter methods are required for spectroscopic properties, they are not adapted to the efficient exploration of conformational landscapes, for which simulations based on force fields remain the method of choice. Moreover, it is not clear whether the electronic structure methods that are appropriate for spectroscopic applications could be straightforwardly transferred to large biomolecules without improving the description of various ingredients such as long-range interactions or anharmonicities. Nevertheless, conformational trends within families of relatively large peptides could be identified by comparing calculated diffusion cross sections for sets of candidate structures provided by molecular mechanics simulations with IMS data.7,13,28 One strategy to get deep insight into the role of individual amino acids on peptide conformation is to study sets of similar systems only differing by specific residue substitutions. This

10.1021/jp102621m  2010 American Chemical Society Published on Web 06/10/2010

Conformation of Polyalanine and Polyglycine Dications

J. Phys. Chem. A, Vol. 114, No. 25, 2010 6889

Figure 1. Sketch of the experimental setup.

idea takes credit from the observation that a single mutation in a sequence can induce conformational changes and alter the biological activity of a peptide or a protein.29 Following such lines, the relation between sequence and conformation has been addressed, for example through the substitution of alanine by other residues at different locations, and by incorporating charged residues in the middle of the sequence.8,30,31 The present work builds upon this effort by systematically investigating how each natural amino acid affects the secondary structure of alanine- and glycine-based peptides. In this purpose ion mobility experiments have been performed together with replica-exchange molecular dynamics simulations. The strategy adopted for structural assignment relies on the comparison of measured and calculated cross section distributions rather than on the bare average values. Two model sequences have been chosen, namely H-Arg(Ala)4X(Ala)4Lys-OH and H-Arg(Gly)4X(Gly)4Lys-OH, where X stands for one of the 20 naturally occurring amino acids. We focus here on doubly protonated peptides, which have recently been studied by electron capture dissociation (ECD).32 Owing to the favored localization of the two protons on each basic residue, this charge state is also more convenient for theoretical modeling. Furthermore, the results from ref 31 demonstrate a direct correlation of amino acid polarity and hydrophobicity with ECD product ion abundances. Therefore, establishing the conformation of doubly charged polyalanine and polyglycine peptides could be important to rationalize this surprisingly linear correlation. The comparison between measured and calculated mobility spectra generally shows good agreement, which highlights the strong potential of this combined experimental/theoretical approach for structural determination. Our results confirm that polyalanines favor helical conformations, except when proline is the central residue. In contrast, polyglycines exhibit a broad variety of secondary structures that are qualitatively captured by the simulations. In addition, the present results shed light on the contribution of the side chains to stabilizing the peptide conformation through charge solvation. The paper is organized as follows. The next section describes the experimental and theoretical methods used to determine the conformations of the peptides. The results are presented and discussed in section III, and concluding comments are given in section IV. II. Methods A. Experimental Section. 1. Peptide Synthesis, Sample Preparation. Standard peptides were purchased from SigmaAldrich (Buchs, Switzerland). All H-Arg(Ala)4X(Ala)4Lys-OH and H-Arg(Gly)4X(Gly)4Lys-OH peptides were produced by solid-state Fmoc chemistry using an Applied Biosystems 433A

synthesizer and further purified by liquid chromatography (Protein and Peptide Synthesis Facility, Biochemistry Department, University of Lausanne, Switzerland). Peptides were dissolved in water to approximately 1 mM concentration and further diluted in a standard spraying solution (H2O/CH3OH 50:50 volume ratio with 0.1% acetic acid) to a final peptide concentration of about 100 µM. 2. Ion Mobility Measurements. Figure 1 shows a schematic of the IMS/MS instrument used for the present studies. The apparatus consists of a homemade 1 m long drift cell coupled to a commercial quadrupole time-of-flight (micro-qTOF, BrukerDaltonics, Bremen, Germany, mass resolution 10 000). Similar coupling has already been reported,33 but with a different drift cell designed for temperature-dependent experiments. In our setup, electrosprayed ions are introduced through a heated capillary interface held at 473 K. These ions are guided through a first ion funnel and accumulated in a cylindrical RF trap.34 Ion packets are periodically injected in the drift tube, across which they travel driven by a uniform electric field. A second ion funnel is used to focus the diffuse ion packet at the exit of the drift tube and to guide them into a vacuum chamber through a 0.7 mm diameter aperture. They are subsequently conveyed into the qTOF instrument by a series of two ion funnels. They travel through a quadrupole and a collision cell before being orthogonally accelerated into a reflectron time-of-flight massanalyzer. The drift tube assembly consists of 101 metal plates. The drift region is operated between 8.3 and 4.7 V cm-1, which corresponds to a drift voltage of 830-470 V. Helium gas is continuously injected at the end of the drift cell at a flow rate of about 140 sccm and the chamber housing the first ion funnel is pumped by a dry rotary pump. Both injection and pumping rates are software-controlled to maintain a constant pressure of about 10 Torr in the drift region. Additionally, the helium flowing from the tube to the injection chamber prevents neutral molecules from entering the drift cell while the electric field pulls the ions through the funnel against the flow. The cylindrical trap is custom-introduced and designed to trap ions at intermediate pressures close to 10 Torr. Transversal confinement of the ions is obtained by applying a RF field (600 Vpp, 5.8 MHz) on a 6 mm diameter ring electrode. Longitudinal confinement is achieved by a dc field created by two cap electrodes distant by 6 mm. The field is kept uniform by covering the cap electrodes with a mesh. The periodic injection in the drift cell proceeds by applying 10 V between the two cap electrodes at a frequency of 3-6 Hz. The time width of the injected ion packet is 1 ms and the orthogonal injection into the TOF mass analyzer is operated at 10 kHz. Synchronization between the ion trap and the TOF MS is provided by a National Instrument clock board

6890

J. Phys. Chem. A, Vol. 114, No. 25, 2010

Albrieux et al.

under control of custom software. Signal from the TOF MS detector is amplified and accumulated in a time-to-digital converter (Acqiris) using the TOF MS orthogonal extraction pulse as a trigger. Mass spectra are eventually recorded as a function of the drift time. Using a drift voltage of 770 V and a 1 ms ion gate leads to an IM resolution of 50. At a given electric field E, the ions travel across the drift tube at a constant speed VD, and the arrival time tD is related to this field through:

tD )

L L ) νD KE

(1)

where K defines the ion mobility. Under the experimental conditions K is inversely proportional to the orientationally averaged diffusion cross section Ω35

K)

1 3 ze 1 × × + 16 N m M

(

) ( 2πkT ) 1/2

1/2

1 Ω

(2)

In eq 2, ze is the charge of the ion, N is the buffer gas density, k is the Boltzmann constant, T is the temperature, and m and M are the masses of the helium atom and the ion, respectively. In practice, the drift time tD is measured for various drift voltages V and the ion mobility is obtained from the slope of tD as a function of V-1. Beyond the mere average value, the distribution of cross sections can be extracted from the distribution of drift times, which can be exploited for separating conformers. However, it should be kept in mind that the broadening of the ion packet by diffusion also contributes to broadening the drift time distributions. The diffusion contribution can be estimated by performing simulations of the propagation of an ion bunch in the tube, thus providing a rough criterion for detecting the presence of different isomers. Finally, the experimental uncertainties on the determined Ω values are estimated to be no more than 1%. B. Numerical Simulations. The conformational landscapes of the [Arg(Ala)4X(Ala)4Lys+2H]2+ and [Arg(Gly)4X(Gly)4Lys+ 2H]2+ peptides have been explored using computational modeling based on the Amber99 force field.36 The numerical protocol consisted first in locating the most stable structures, then in performing equilibrium simulations at several temperatures of interest. Both tasks rely on replica-exchange molecular dynamics (REMD) simulations, which provide a much more reliable and unbiased approach than simulated annealing.37 In the optimization stage, 40 trajectories were conducted in parallel and followed by systematic quenching of a pool of 1000 configurations recorded periodically. The molecular dynamics simulations (MD) were carried out with a dielectric constant ε ) 1 and a time step of 1 fs. Thermalization was achieved using a Berendsen thermostat with coupling constant of 1 ps-1. A random exchange between a pair of configurations from adjacent replicas was attempted every 100 fs, and the temperatures were allocated according to a geometric progression in the temperature range 80-1000 K. The MD simulations for each trajectory were propagated for 1 ns and initiated from fully extended conformations, in which the backbone torsions are set to 180°. After conjugate gradient minimization of the 40 000 recorded configurations, the REMD were restarted by initiating all replicas in the lowest-energy structure. A second series of quenches sometimes resulted in improving over these conformations, in which case a third REMD simulation was carried out. No need for a fourth iteration of this procedure turned out to be necessary,

the most stable structure found at the third stage being never lower in energy than the previously determined global minimum. Once the putative global minima had been located, an additional REMD simulation was carried out to determine thermal equilibrium properties. Only 7 replicas were used, distributed according to an arithmetic progression in the temperature range 100-500 K but with improved statistics of 10 ns for each replica. Again, representative samples of 1000 configurations per replica were saved. Their cross sections were calculated for the 300 K replica using the direct trajectory method of Mesleh and co-workers.38 It is important to emphasize that the calculated distribution of diffusion cross sections reflects the distribution of structures explored at 300 K during the 10 ns MD simulation, while the experimental distributions are also affected by the natural broadening due to diffusion. For the structure samples, a statistical analysis of the conformations was conducted by calculating the couples of associated (φ, ψ) torsion angles along the backbone chain, thus providing convenient representations as Ramachandran scatter plots. The simulations were performed for all the experimentally investigated systems. For most amino acids except X ) Arg, His, and Lys, the terminal Arg and Lys residues are expected to carry the two additional protons. However, for the mentioned exceptions, it is unclear which two residues among the three basic ones would be protonated. For such cases several possibilities were considered, resulting from the assumption that the nonprotonated residue was either the central one or the C-terminal lysine chain, the N-terminal arginine always being charged. This yields two possible protonation sites for each case. A subcase arises for neutral central His, for which the two tautomeric forms (two possible binding sites for the remaining proton on the His side chain) were independently treated. Making the assumption of fixed protonation sites may be oversimplified due to the neglect of possible proton transfer between the basic groups.39,40 Stable complexes with delocalized protons may even be formed and influence the finite-temperature behavior.41 However, a proper treatment of proton transfer for such doubly protonated species would require a significantly more advanced modeling, including reactivity, which lies beyond the present computational capabilities. III. Results A. Alanine-Based Peptides. Figure 2 shows the distributions of diffusion cross sections measured for all [Arg(Ala)4X(Ala)4Lys+2H]2+ peptides, where X covers the entire series of naturally occurring amino acids. All distributions range between 230 and 250 Å2, and generally consist of a single peak. The width of these peaks is compatible with natural diffusion, which suggests that mainly one conformer is observed in most cases. However, a bimodal distribution is observed for X ) Lys and, to a lesser extent for X ) Glu and Cys. The calculated cross-section distributions are superimposed to the experimental data in Figure 2 for the entire series of peptides, including several protonation sites or tautomeric forms for Arg, Lys, and His. Regarding the position of the peaks, the agreement between experimental and calculated distributions is quantitative for most sequences, which shows that the conformational sampling and the present force field yield realistic structures. Regarding the widths, the REMD simulations at 300 K also turn out to reproduce the distributions very satifactorily. The most stable structures obtained for selected cases, namely the Ala-, Asp-, Cys-, and Pro-containing peptides, are depicted

Conformation of Polyalanine and Polyglycine Dications

J. Phys. Chem. A, Vol. 114, No. 25, 2010 6891

Figure 2. Diffusion cross section distributions for [Arg-(Ala)4-X(Ala)4-Lys+2H]2+ as a function of the central residue X. Experimental data are in gray lines. Distributions calculated by assuming protonation at Arg1 and Lys11 are plotted as black lines; black dashed lines stand for protonation of Arg1 and X6. In the case of His, the distributions obtained for the two possible neutral tautomers are depicted by the black line and the red dashed line, respectively.

in Figure 3. In this figure we also show the corresponding Ramachandran diagrams obtained by plotting the (φ, ψ) couples extracted from 1000 structures taken from the 300 K trajectory. In these plots, the backbone dihedral angles for all the residues are distinguished from the sole contribution of the central residue. Similar data for the other sequences are given as Supporting Information. Except for X ) Pro and structures with one charge located in the middle of the sequence (see below), all conformations are based on an R-helical secondary structure which includes the central residue. This trend is confirmed on the corresponding Ramachandran plot, which is dominated by a spot in the R-helical region. Features common to all these sequences include an attractive electrostatic interaction between the positively charged arginine side chain and the COOH terminal group, as well as the solvation of the lysine charge by the peptide bond carbonyl groups that are located at the C-terminal moiety of the peptide. This latter motif, which has already been reported for singly protonated alanine-based peptides,7 shows that the helical structure is stabilized due to a favorable electrostatic interaction between the charge and the helix electric dipole. Beyond these common features, this family of helical conformations can be divided into three main groups depending on the central residue X. The presence of a carboxyl group in the side chain of X, or at least an oxygen atom, favors an electrostatic attraction with the charge on Arg. This is observed for X ) Asp (displayed in Figure 3) and for X ) Glu, for which the two carboxyl groups bind to the Arg side chain. Concomitantly, the helical structure partially unfolds near both the C-terminal and N-terminal ends, which appears on the Ramachandran plots as intense spots around (-75°, +50°) and (70°, 170°), thus away from the R-helical region. Though less pronounced, the same effect is also observed for X ) Gln, Asn,

Figure 3. Ramachandran plots for selected [Arg-(Ala)4-X-(Ala)4Lys+2H]2+ extracted from samples of 1000 structures obtained from the 300 K REMD trajectories. The (φ, ψ) angles for the central residue are plotted as black dots and superimposed over a density plot of the backbone dihedral angles for all residues. The most stable structures are displayed next to the plot.

and Thr, due to the oxygen in their side chain. In these peptides the helix is rather short and ranges from residue 3-4 to residue 7. A second group is formed by the Trp-, Met-, Phe-, Tyr-, Ile-, and Val-containing peptides, for which no significant interaction is found between the side chain of X and the charge on Arg. Their lowest-energy structures are helical from residue 3 up to residue 10, i.e., over the major part of the backbone. This can be seen on the Ramachandran plot for X ) Trp where most (φ, ψ) couples appear in the R-helical region. In the third group (X ) Ala, Cys, Gly), no charge solvation involving the side chain of X is obtained and the solvation scheme of the lysine side chain by the backbone carbonyl group is preserved. However, compared to the case for Trp, the unfolding on the N-terminal side is more pronounced in this group. As a consequence, helices with intermediate length (over residues 5-10) are formed, leading to Ramachandran plots that are similar to those of the Trp group. The difference between the peptides from the Asp group and those from the Trp group can be understood as the result of a competition between optimal

6892

J. Phys. Chem. A, Vol. 114, No. 25, 2010

Albrieux et al.

Figure 4. Diffusion cross section distributions calculated for Arg(Ala)4-Arg-(Ala)4-Lys assuming two different protonation schemes: (a) on both Arg1 and Lys11 (solid line); (b) on Arg1 and Arg6 (dashed line).

charge solvation of the side chains and helix stabilization through H-bond network formation. This does not apply to the Ala group because the central side chains are too small to destabilize the helix. The small size of the Ser side chain may thus be also responsible for its apparent similarity with Ala. The particular behavior of the smaller member of the Asp group, X ) Thr, might give some support to this last statement. Two conformers are found to coexist in the 300 K trajectory calculated for this peptide (see Supporting Information). The most stable structure shares the common characteristics of the Asp group, while the other populated isomer is similar to X ) Ala. Finally, X ) Leu turns out to adopt a conformation intermediate between X ) Asp and X ) Trp, with a short helical domain, from residue 4 to residue 7, and no charge solvation by the side chain of Leu. We now turn to X ) Arg, Lys, and His, for which the protonation sites are ambiguous, and in the case of neutral His, two tautomeric forms are available. The results of the simulations performed assuming fixed protonation sites or tautomeric form are shown in Figure 2. Unfortunately, comparison with experimental data does not discriminate between those different assumptions, even though the structures obtained show major differences. Figure 4 illustrates the case of X ) Arg, which in neutral form yields a helical structure intermediate between X ) Trp and X ) Asp. In contrast, assuming that the central Arg is protonated with a neutralized terminal Lys results in a poorly ordered conformer. However, both structures roughly show the same cross section distributions, at least with respect to the experimental resolution. The same charge effect on the structure is observed for Lys and His, but the shape and the positions of the peaks are better resolved. For Lys, the bimodality observed in the experimental distribution does not seem compatible with the two calculated conformers. Although possible, such coexistence is also difficult to assess in the other two systems due to overlapping cross-section distributions. The structure obtained for X ) Pro (see Figure 3) is essentially globular and the small measured cross section is well reproduced by the simulations. The high rigidity of proline is reflected by the value of its φ angle ranging around -75°, whereas the Ψ angle is more flexible. As a consequence, the geometry of the closest alanine residues is constrained and those appear as an intense spot around (φ, ψ) ) (-70°, +140°). Despite this local disturbance, the torsion angles for most other Ala residues lie within the R-helical region of the plot. On the other hand, the solvation scheme observed for the other peptides is preserved, with the charges on Arg and Lys side chains solvated by COOH and the backbone carbonyl groups, respectively. B. Glycine-Based Peptides. Figure 5 shows the distributions of diffusion cross sections measured for all [Arg(Gly)4X(Gly)4-

Figure 5. Diffusion cross section distributions for [Arg-(Gly)4-X(Gly)4-Lys+2H]2+ as a function of the central residue X. Experimental data are in gray lines. Distributions calculated by assuming protonation at Arg1 and Lys11 are plotted as black lines; black dashed lines stand for protonation of Arg1 and X6. In the case of His, the distributions obtained for the two possible neutral tautomers are depicted by the black line and the red dashed line, respectively.

Lys+2H]2+ peptides except X ) Cys, for which signal was not sufficient. The peptides from this second series all have cross sections ranging from 200 to 230 Å2 and are thus smaller than the polyalanines, as expected due to the smaller size of glycine. Most distributions are dominated by a single peak, and their width is compatible with the natural diffusion of a single conformer. However, for X ) His, the distribution is sufficiently large to host multiple structures. The cross section distributions obtained from REMD simulations, also displayed in Figure 5 for all peptides, are generally singly peaked except for X ) Phe, which clearly exhibits some bimodal character. The widths of the measured distributions are generally well reproduced, even for X ) His, if we assume that the terminal Lys is neutral. Only X ) Ala shows a calculated distribution much broader than in the experiment. Regarding the peaks position, the agreement is very satisfactory for most peptides but the small X ) Ala, Gly, Ser, Asp, and Arg, for which the calculated cross sections are systematically underestimated. Unlike the previously discussed polyalanines, where the stable conformations could be characterized in terms of their helical content, no such sorting seems obvious for polyglycines. Some global tendencies can nevertheless be inferred from representative examples. Following the same protocol as for alanine-based peptides, Ramachandran plots were extracted from structures sampled at 300 K. Such plots are shown in Figure 6 for X ) Gly, Thr, Ile, and Phe, together with the corresponding lowest-energy structures. In some cases, where several stable conformers could be identified from the equilibrium simulations, other relevant structures are depicted (data for other amino acids are given as Supporting Information). Compared to those for polyalanines, all Ramachandran diagrams of polyglycines

Conformation of Polyalanine and Polyglycine Dications

J. Phys. Chem. A, Vol. 114, No. 25, 2010 6893

Figure 6. Ramachandran plots for selected [Arg-(Gly)4-X-(Gly)4-Lys+2H]2+ extracted from samples of 1000 structures obtained from the 300 K REMD trajectories. The (φ, ψ) angles for the central residue are plotted as black dots and superimposed over a density plot of the backbone dihedral angles for all residues. The most stable structures found for X ) Gly and Thr are displayed next to the corresponding plots. For X ) Phe and Ile, two coexisting structures are depicted with the lowest-energy conformation labeled as (a).

display broader ranges of accessible dihedral angles, which reflects the higher flexibility of glycine-based backbones. Moreover, the diagrams are significantly altered upon substituting the central amino acid. The large variety of conformations revealed by the Ramachandran plots may be interpreted as the absence of dominant secondary structure but may also be due to the coexistence of multiple isomers. Such effects can be disentangled by looking at the local (φ, ψ) dihedrals of the central amino acid. In nearly half the cases, including X ) Gly and Thr (see Figure 6), these conformations are localized in very narrow regions. Further analysis of the trajectories confirms that each of these peptides adopts a single well-defined conformation similar to the lowestenergy structure (within thermal fluctuations). Visual inspection indicates that charge solvation is the most important stabilizing factor, in particular near the peptide bond carbonyl group and the central residue side chain. In contrast, the remaining half of the peptide series exhibits some signature of isomer coexistence. This coexistence is manifested on the bimodal (X ) Phe) or particularly broad

(X ) His) cross section distributions in Figure 5. For other peptides, the existence of various conformers is evidenced on the Ramachandran plots. For instance, the (φ, ψ) angles for Ile are distributed in two distinct spots (see Figure 6). The spot centered at (φ, ψ) ) (-150°, 0°), which includes the most stable conformation, has the Arg charged side chain parallel to the terminal COOH and on top of it. On the other hand, these two charged groups lie in the same plane in conformations belonging to the second spot at (φ, ψ) ) (-60°, -20°). Both isomers are illustrated in Figure 6. Similar behavior is found for X ) Gln, Leu, Met, and Asn, as well as for X ) Arg and His, if we assume the central residue to be protonated. In all these cases, the two main conformers suggested by the bimodal Ramachandran diagram differ in the local arrangement and solvation of the charges and side chains. Finally, X ) Ala and one of the tautomers for X ) His display three resolved spots. In the case of Ala, the corresponding conformers share a common solvation pattern of the carboxyl end and the Arg side chain but differ in the solvation of the Lys side chain by the peptide bond carbonyl groups. In the case of His, the differences lie in the relative

6894

J. Phys. Chem. A, Vol. 114, No. 25, 2010

Figure 7. Average diffusion cross section of [Arg-(Ala)4-X-(Ala)4Lys+2H]2+ as a function of the volume of the central amino acid from Counterman et al.43 (black squares). The nature of the central amino acid is labeled using the one-letter code. For X ) Lys, the measured cross section distribution is bimodal and the two corresponding peaks are denoted as K and K*. Linear fitting of the data except Pro and K* (solid line) provides an estimate for the volumes of Cys, His, and Arg (red triangles). The same correlation plot for [Arg-(Gly)4-X-(Gly)4Lys+2H]2+ is shown in the inset.

position of the side chains of His and of the charged terminal residues. Note that the existence of several spots on the Ramachandran plot is not necessarily related with the coexistence of clearly distinct conformers. The two spots for X ) Glu denote two possible local arrangements of the main chain without global structural change, especially in the charge solvation pattern. C. Discussion. The different conformations of the present alanine- and glycine-based peptides are found to be mainly determined by charge solvation of the Arg side chain, carboxyl group and, albeit to a lesser extent, the side chain of the central residue. This is particularly striking in polyglycines, for which the many possible ways of solvating the charged groups are facilitated by the flexibility of the glycine amino acid, also reflected on the broad variety of structures sampled at 300 K. The stronger propensity of Ala to form helices leads to a common motif in all polyalanines except X ) Pro, differences between these peptides originating from the contrasted participation of the central side chain to charge solvation. As a consequence, the extent of helical order varies with the central residue. Interestingly, and although the proline-based peptide does not show helical conformations, this dependence is poorly correlated with known helix propensity scales42 for other amino acids. For instance, the helical length is 7 for X ) Trp but only 5 for X ) Glu, whereas glutamic-acid is expected to favor helices better than tryptophan. The rather similar secondary structures obtained for most of the investigated polyalanines suggests that the differences in the measured cross sections may be related to the geometrical extent of the central amino acid. The volume V of individual amino acids has been reported by Counterman et al.,43 except for Arg, Cys, and His. Figure 7 shows the average cross sections measured for all peptides as a function of V. In the bimodal case of Lys, the less pronounced peak is highlighted as K*. Except for proline (P) and K*, the measured cross-section increases roughly linearly with V. This linear behavior allows the volumes missing for Arg, Cys, and His to be estimated. Excluding P and K* from the fit, we find V(Arg) ) 130 Å3, V(Cys) ) 90 Å3, and V(His) ) 125 Å3, with an estimated uncertainty of about 30%.

Albrieux et al.

Figure 8. Potential energy versus diffusion cross section for a series of 1000 conformations of (a) [Arg-(Ala)4-Ile-(Ala)4-Lys+2H]2+ and (b) [Arg-(Gly)4-Ile-(Gly)4-Lys+2H]2+, sampled at 300 K. Instantaneous and optimized structures are denoted as black crosses and red circles, respectively.

The correlation is much less pronounced for the polyglycines (see inset of Figure 7), as expected from the spreading of measured cross sections. It should be noted that central residues that are chemically similar can produce markedly different conformers, as illustrated on the cross section distributions measured for X ) Asp and X ) Glu. Likewise, the distributions obtained for X ) Tyr, Trp, and Phe, although experimentally similar to each other, show more compact structures for the latter amino acid in the simulations. For this peptide, agreement is obtained only with a secondary peak in the cross section distribution, which indicates metastable conformers. In cases where the experimental distributions are well reproduced by the simulations, the structures sampled at 300 K can be analyzed further by mapping their cross sections to their energies. We have represented in Figure 8 the correlation between cross section and instantaneous potential energy obtained for the 1000 representative structures of the [Arg(Ala)4Ile(Ala)4Lys+2H]2+ and [Arg(Gly)4Ile(Gly)4Lys+ 2H]2+ peptides. From an energy landscape perspective, we have also locally optimized these structures, and reported their cross section versus energy relation in the same figure. Interestingly, while glycine is more flexible than alanine and produces a bimodal distribution of (φ, ψ) angles, fewer local minima are obtained from the 1000 instantaneous structures, namely 231 instead of 768 for the polyalanine. The larger number of distinct inherent structures reflects the numerous slightly different arrangements of the methyl side chains. The cross sections and instantaneous potential energies are scattered over one main region in each diagram, as expected from the singly peaked distributions of Figures 2 and 5. Upon local optimization, the energies drop by more than 100 kcal/mol and become spread in much narrower ranges. The diffusion cross sections do not vary much after optimization but drop by about 1 and 2 Å2 in polyalanines and polyglycines, respectively. Such a decrease in the cross section of locally optimized structures was expected, since vibrational thermalization leads to an expansion in the system. However, this effect lies below the experimental resolution. Finally, the repartition of the cross sections with respect to the corresponding energies for the inherent structures is more scattered and uniform in polyglycines than in polyalanines, where lower energy structures converge toward an approximately constant cross section of 240 Å2. This also correlates with the broader variety of conformations in the former system.

Conformation of Polyalanine and Polyglycine Dications Deviations between the measured cross sections and the calculated structures, though minor in polyalanines and in half the polyglycines, are more salient in the latter cases when small or basic central amino acids are involved, as well as Asp. Such discrepancies are likely due to the oversimplified treatment of electrostatic interactions and the neglect of proton migration in the presently used Amber force field. More sophisticated models for multiply charged peptides should at least account for polarization forces. However, the possible influence of proton transfer remains unclear. The present calculations clearly indicate that several topologically different conformers may be compatible with the measured cross section distribution, even when the latter are not particularly broad (as for X ) Ile). The peptides with competing residues (X ) Arg, Lys), which do not show broad distributions either, may thus correspond to one or several conformers sharing similar diffusion cross sections. Proton migration may play various roles in this context. The two protons may be fixed on two basic sites among the three available acceptor groups, giving well-defined stable conformers, but the finite temperature may also favor statistical coexistence between more than one isomer. But proton transfer could have a more active chemical influence by stabilizing proton-bridged complexes similar to the Zundel cation (H5O2+), thereby leading again to single well-defined conformers. Theoretical evidence for such complexes was previously reported for shorter peptides.41 In the case of X ) His, and in addition to the possible influence of proton transfer for this rather basic amino acid, the existence of the two tautomeric forms of neutral histidin may further contribute to explaining why the cross section distribution is particularly broad. IV. Conclusions Even without surrounding solvent, peptides in the gas phase tend to show a broad variety of structures. In the present work, we have attempted to characterize the three-dimensional conformation of the series of alanine- and glycine-based peptides [Arg(Ala)4X(Ala)4Lys+2H]2+ and [Arg(Gly)4X(Gly)4Lys+ 2H]2+, X covering the entire range of natural amino acids. In this purpose, ion mobility spectrometry measurements were carried out and combined with replica-exchange molecular dynamics simulations. All polyalanines except the one containing Pro are found to be based on the R-helix motif, differences being due to slightly different arrangements of the central side chain relatively to the charged arginine side chain and the carboxyl group. The common trends displayed by polyalanines are further manifested on the nearly linear relation between the diffusion cross section and the volume of the central amino acid. In contrast, polyglycines exhibit a much broader variety of conformations and a significant dependence on the central residue that both reflect the known greater flexibility of glycine with respect to the helix-former alanine. Charge solvation was identified as the main cause explaining the conformation of polyglycines, again involving the arginine side chain, the carboxyl end group, and a greater influence of the central amino acid side chain. Polyglycines also often show a rugged energy landscape under the form of multimodal stability between several conformers, in the simulations but also on broader cross section distributions in the experiment. However, one outcome of the present work is that structures may appear in competition on Ramachandran diagrams without having clearly resolved diffusion cross sections. While the present theoretical results satisfactorily agree with the measurements in most cases, polyglycines seem to pose a greater challenge to computation. The force field chosen here

J. Phys. Chem. A, Vol. 114, No. 25, 2010 6895 does not account for polarization interactions, which are probably important for doubly protonated peptides. In addition, the possible intramolecular proton mobility in cases where the central amino acid is basic (Arg, Lys, His) would require additional modeling, to explain why some peptides show particularly broad cross section distributions. One possible way of improving the agreement between experimental and calculated structures would be to bias sampling toward specific values of the cross section, using for instance umbrella sampling. Rather than enforcing the bias on the cross section directly (which would imply a significant computational overcost), such guiding could be achieved by targetting regions of configuration space with geometric properties correlated to the cross section, primarily the gyration radius but also higher-order deformation parameters. Finally, it should be noted that most of our structural analysis relied on the lowest-energy conformers. Those may not accurately represent the energy landscapes at room temperature, especially in the case of multimodality. It would then be valuable to characterize the peptides in terms of their free energy, but this more ambitious task lies beyond the scope of the present work. Despite such technical improvements, the combination of ion mobility spectrometry with molecular simulation already turns out as a powerful combination for unraveling the conformation of peptides in the gas phase and, more importantly, the physical and chemical factors responsible for those conformations. The systematic series of peptides investigated here not only provide a suitable testing case for this methodology but also should be valuable in future experiments now that their main structural features have been elucidated. In particular, the knowledge of structure paves the way for new conformation-resolved experiments, where the influence of secondary motifs on various properties will become accessible. Particularly, we can now proceed to rationalize the influence of peptide structure on product ion abundance distributions obtained in ECD experiments.32 Acknowledgment. The IM instrument was built with funding from Agence Nationale pour la Recherche (projet MS+). Y.T. and A.V. appreciate financial support of the Swiss National Science Foundation (SNF project 200021-125147/1) and the Swiss Academy for Engineering Sciences (SATW project 200928). Supporting Information Available: Ramachandran plots for all [Arg-(Ala)4-X-(Ala)4-Lys+2H]2+ and [Arg-(Gly)4-X(Gly)4-Lys+2H]2+ peptides with corresponding structures. This material is available free of charge via the Internet at http:// pubs.acs.org. References and Notes (1) Fersht, A. Structure and mechanisms in protein science; Freeman: New York, 1999. (2) Dill, K. A. Biochemistry 1990, 29, 7133–7155. (3) Jarrold, M. F. Annu. ReV. Phys. Chem. 2000, 51, 179–207. (4) Barran, P. E.; Polfer, N. C.; Campopiano, D. J.; Clarke, D. J.; Langridge-Smith, P. R.; Langley, R. J.; Govan, J. R.; Maxwell, A.; Dorin, J. R.; Millar, R. P.; Bowers, M. T. Int. J. Mass Spectrom. 2005, 240, 273– 284. (5) Themed issue on Biomolecular structures: from isolated molecules to living cells. Physical Chemisty Chemical Physics; Kim, S. K., Ha, T., Schermann, J. P., Eds.; Royal Society of Chemistry: London, 2010; Vol. 12, pp 3317-3632. (6) Counterman, A. E.; Clemmer, D. E. J. Am. Chem. Soc. 2001, 123, 1490–1498. (7) Hudgins, R. R.; Ratner, M. A.; Jarrold, M. F. J. Am. Chem. Soc. 1998, 120, 12974–12975. (8) Jarrold, M. F. Phys. Chem. Chem. Phys. 2007, 9, 1659–1671.

6896

J. Phys. Chem. A, Vol. 114, No. 25, 2010

(9) Chou, P.; Fasman, G. Biochemistry 1974, 13, 211–222. (10) Richardson, J.; Richardson, D. Science 1988, 240, 1648–1652. (11) Antoine, R.; Compagnon, I.; Rayane, D.; Broyer, M.; Dugourd, P.; Breaux, G.; Hagemeister, F. C.; Pippen, D.; Hudgins, R. R.; Jarrold, M. F. J. Am. Chem. Soc. 2002, 124, 6737–6741. (12) Hudgins, R. R.; Mao, Y.; Ratner, M. A.; Jarrold, M. F. Biophys. J. 1999, 76, 1591–1597. (13) Wyttenbach, T.; Bushnell, J. E.; Bowers, M. T. J. Am. Chem. Soc. 1998, 120, 5098–5103. (14) Liu, D.; Wyttenbach, T.; Bowers, M. T. Int. J. Mass Spectrom. 2004, 236, 81–90. (15) Dugourd, P.; Antoine, R.; Breaux, G.; Broyer, M.; Jarrold, M. F. J. Am. Chem. Soc. 2005, 127, 4675–4679. (16) Laskin, J.; Futrell, J. H. Mass Spectrom. ReV. 2003, 22, 158–181. (17) Reilly, J. P. Mass Spectrom. ReV. 2009, 28, 425–447. (18) Zubarev, R. A. Mass Spectrom. ReV. 2003, 22, 57–77. (19) Cooper, H. J.; Hakansson, K.; Marshall, A. G. Mass Spectrom. ReV. 2005, 24, 201–222. (20) Gabelica, V.; Tabarin, T.; Antoine, R.; Rosu, F.; Compagnon, I.; Broyer, M.; De Pauw, E.; Dugourd, P. Anal. Chem. 2006, 78, 6564–6572. (21) Larraillet, V.; Antoine, R.; Dugourd, P.; Lemoine, J. Anal. Chem. 2009, 81, 8410–8416. (22) Englander, S. W. J. Am. Soc. Mass Spectrom. 2006, 17, 1481– 1489. (23) Chin, W.; Piuzzi, F.; Dimicoli, I.; Mons, M. Phys. Chem. Chem. Phys. 2006, 8, 1033–1048. (24) Abo-Riziq, A.; Crews, B. O.; Callahan, M. P.; Grace, L.; de Vries, M. S. Angew. Chem., Int. Ed. 2006, 45, 5166–5169. (25) Stearns, J. A.; Seaiby, C.; Boyarkin, O. V.; Rizzo, T. R. Phys. Chem. Chem. Phys. 2009, 11, 125–132. (26) Clemmer, D. E.; Hudgins, R. R.; Jarrold, M. F. J. Am. Chem. Soc. 1995, 117, 10141–10142.

Albrieux et al. (27) von Helden, G.; Wyttenbach, T.; Bowers, M. T. Science 1995, 267, 1483–1485. (28) Fernandez-Lima, F. A.; Wei, H.; Gao, Y. Q.; Russell, D. H. J. Phys. Chem. A 2009, 113, 8221–8234. (29) Creighton, T. E. Proteins: Structures and Molecular Properties; Freeman: New York, 1984. (30) Srebalus Barnes, C. A.; Clemmer, D. E. J. Phys. Chem. A 2003, 107, 10566–10579. (31) McLean, J. R.; McLean, J. A.; Wu, Z.; Becker, C.; Peı`rez, L. M.; Pace, C. N.; Scholtz, J. M.; Russell, D. H. J. Phys. Chem. B 2010, 114, 809–816. (32) Vorobyev, A.; Hamidane, H. B.; Tsybin, Y. O. J. Am. Soc. Mass Spectrom. 2009, 20, 2273–2283. (33) McCullough, B. J.; Kalapothakis, J.; Eastwood, H.; Kemper, P.; MacMillan, D.; Taylor, K.; Dorin, J. R.; Barran, P. E. Anal. Chem. 2008, 80, 6336–6344. (34) Albrieux, F.; Antoine, R.; Chirot, F.; Lemoine, J.; Dugourd, P. Submitted for publication. (35) Revercomb, H. E.; Mason, E. A. Anal. Chem. 1975, 47, 970–983. (36) Wang, J.; Cieplak, P.; Kollman, P. J. Comput. Chem. 2000, 21, 1074–1049. (37) Wales, D. J.; Scheraga, H. A. Science 1999, 285, 1368–1372. (38) Mesleh, M. F.; Hunter, J. M.; Shvartsburg, A. A.; Schatz, G. C.; Jarrold, M. F. J. Phys. Chem. 1996, 100, 16082–16086. (39) Kohtani, M.; Jones, T. C.; Sudha, R.; Jarrold, M. F. J. Am. Chem. Soc. 2006, 128, 7193–7197. (40) Maksic, Z. B.; Kovacevic, B. Chem. Phys. Lett. 1999, 307, 497– 504. (41) Calvo, F.; Dugourd, P. J. Phys. Chem. A 2008, 112, 4679–4687. (42) Pace, N. C.; Scholtz, M. J. Biophys. J. 1998, 75, 422–427. (43) Counterman, A. E.; Clemmer, D. E. J. Am. Chem. Soc. 1999, 121, 4031–4039.

JP102621M