ARTICLE pubs.acs.org/JPCB
Structural Properties of a Membrane Associated Anchor Dipeptide Victor V. Volkov,† Riccardo Chelli,†,‡ Francesco Muniz-Miranda,† and Roberto Righini*,†,‡ † ‡
European Laboratory for Nonlinear Spectroscopy (LENS), Universita di Firenze, Via Nello Carrara 1, I-50019 Sesto Fiorentino, Italy Dipartimento di Chimica “Ugo Schiff”, Universita di Firenze, Via della Lastruccia 3, I-50019 Sesto Fiorentino, Italy
bS Supporting Information ABSTRACT: The association of peptides to phospholipid membranes through the insertion of an anchoring hydrocarbon tail is common to some viruses and to several anticancer drugs. We investigate the association of an anchor dipeptide, Nmyristoylated methyl glycine (MrG), to phospholipid membrane fragments made of 1-palmitoyl-2-linoleyl phosphatidylcholine (PLPC). Here we report on the experimental findings of two-dimensional infrared spectroscopy of an MrG backbone in the 6 μm wavelength region. The experimental outcomes are supported by ab initio calculations and by a molecular dynamics simulation accomplished with the replica exchange method. We find that the guest molecule has a preferential unfolded conformation, with dihedral angles Φ = 90 ( 20° and Ψ = 180 ( 20°, while the average orientational distribution of the amide I transition dipole moments with respect to the neighbor PLPC carbonyls is peaked at angles in the range 2133°. The depth of penetration of MrG inside the membrane corresponds rather well to the one estimated in our previous paper [J. Phys. Chem. B, 2009, 113, 16246], where we found that the backbone moieties of MrG are localized slightly above the carbonyl groups of PLPC. According to the simulation results, the anchor tail is completely inserted in the hydrophobic region of the bilayer, with a largely prevalent extended conformation and a preferential alignment along the average direction of the PLPC hydrocarbon tails.
’ INTRODUCTION Small polypeptides, for the relative simplicity of their primary sequences and their healing propensity, are attractive subjects in drug design studies and in clinical treatments of several pathologies, in particular of various forms of cancer.14 Even though the knowledge of the structural and dynamic properties of potentially interesting sequences is of great importance in anticipating their functional capabilities, the selection process of substances for medical applications is essentially based on clinical evaluations and tests, which are mostly phenomenological in character. Furthermore, even for certified drugs of this type admitted to clinical practice, the knowledge of their structural properties is often very limited. One of the reasons is that, when injected into blood plasma, the molecule experiences a variety of different environments, including possible partition into the phospholipid membrane. Optical spectroscopy is rather unique as a tool for identifying the structural properties of molecules in disordered phospholipid membrane environment. In fact, approaches based on the measurement of macroscopic properties, such as calorimetric and electrochemical techniques, can hardly help in the identification of the relative arrangement of molecular moieties on a local scale, while X-ray and electron scattering experiments have limited application due to the disorder of the membrane samples in physiologic conditions. In these systems, NMR spectra are heavily broadened as a result of the slow molecular motions on r 2011 American Chemical Society
the time scale of the sampling process. On the contrary, contemporary methods of nonlinear optical spectroscopy make it possible to probe both the intra- and intermolecular structural correlations between chromophores, including their temporal evolution on a subpicosecond time scale. Most commonly, third order nonlinear experiments in the infrared57 and the visible810 regions allow probing certain components of the molecular susceptibility and, hence, the related structural properties on the time scale of the atomic motions. Infrared (IR) multidimensional spectroscopy, being sensitive to local properties, is particularly helpful in molecular structural analysis. In particular, two-dimensional infrared (2D-IR) spectroscopy was recently applied in the characterization of phospholipid membranes11 and in structural and dynamical analysis of several polypeptides (including those of potential anticancer activity) associated with phospholipid bilayers.1214 Even though multidimensional spectroscopy is under rapid development at present, one has to keep in mind that the resolving power of 2D-IR spectroscopies is, in some cases, limited by persisting ambiguities due, for instance, to the coexistence of several molecular conformations.15 It has been demonstrated that the Special Issue: Shaul Mukamel Festschrift Received: September 28, 2010 Revised: February 15, 2011 Published: March 28, 2011 5294
dx.doi.org/10.1021/jp109284z | J. Phys. Chem. B 2011, 115, 5294–5303
The Journal of Physical Chemistry B combination of nonlinear spectroscopic experiments and molecular dynamics (MD) simulations makes it possible to take advantage of the fine features of the 2D-IR spectral dispersion to get information on the structural variance of the molecular system under study.1517 Additionally, as demonstrated recently, an increase in dimensionality (up to fifth order) helps to improve the resolving power of the spectroscopic experiments.18 In this article we combine methods of linear and nonlinear infrared spectroscopy with classical MD simulations and quantum mechanical calculations to learn how far we can progress in the structural characterization of a simple dipeptide such as N-myristoylated-methyl-glycine (MrG) in interaction with phospholipid membrane fragments. The dipeptide is designed to mimic the natural anchor found in a variety of membrane associated polypeptides.1923 There are several possible forms of association with a bilayer. Some proteins, such as G-proteins and porins, play both a constituent and a functional role, penetrating the bilayer as integrated helical-cage systems or barrel-like structures, and at the same time carrying important functions of ion transport and/or signal transduction.18 Other proteins may associate with the membrane temporarily through electrostatic and hydrophobic interactions with specific chemical moieties at the phospholipid membrane interface.1921 In a third format, the association sets up by means of hydrophobic tails (the anchors) attached to certain conservative structural segments of the polypeptide.1923 In this case, anchoring provides an opportunity for a stable, specific and controllable form of interaction of the polypeptide with the membrane surface (of certain composition and fluidity) on the time scale of cellular metabolism. With its myristoyl tail, MrG represents a typical example of association to membranes through this third mechanism. Understanding the details of this peptidemembrane interaction at the molecular level, beside the general relevance of learning about the effective anchoring of proteins to phospholipid bilayers, recently became a critical step in the realization of novel anticancer remedies.24 In a previous paper14 we used two-color 2D-IR spectroscopy to determine the depth of penetration of the MrG backbone into the polar fraction of the phospholipid bilayer made of 1-palmitoyl2-linoleyl phosphatidylcholine (PLPC) molecules. The experiment consisted in exciting the water stretching modes and detecting the induced perturbations in the spectral region of the carbonyl vibrations. Here we specifically address the MrG backbone conformation and its localization in the PLPC bilayer, by extending the experimental work to the one-color 2D-IR spectra in the frequency region of the amide I modes. With the essential support of MD simulations and quantum mechanical calculations, we show that the one- and two-color 2D-IR spectroscopic results allow us to attain a fairly exhaustive picture of the structure of the guest molecule, of its interaction with the neighboring moieties and of the local restructuring of the host bilayer.
’ MATERIALS AND METHODS Sample Preparation. The MrG dipeptide (see Figure 1) is synthesized by BIOSYNTAN GmbH, Berlin, Germany, using 13 C isotope labeled myristic acid. Figure 1 shows the position of the carbon isotope in the structure of MrG. The purity of dipeptide is better than 95%. We remove the trifluoroacetic acid (TFA; left after reverse-phase high-performance liquid chromatography (HPLC)) by lyophilizing three times the substance under low pH (pD) condition. Soy L-R-phosphatidylcholine 95%
ARTICLE
Figure 1. Structures of PLPC phospholipid and MrG dipeptide molecules.
extract is from Avanti Polar Lipids (product number 441601G). PLPC (see Figure 1) is the main molecular component of the extract. In order to prepare membrane fragments, first we mechanically mix PLPC and MrG powder in the ratio 10:1. Second, we dissolve the mixture in chloroform and dry it on a glass plate. Then, after addition of D2O, the suspension undergoes mechanical mixing at 60 °C until the sample (50 μm film between two CaF2 windows) demonstrates proper optical quality. The sample suspension undergoes proper aging (2 weeks) to secure good admixing of MrG molecules into the PLPC host environment. Infrared Measurements. The FTIR spectra (Shimadzu 8400S) of the sample suspensions in D2O in the OH stretching region show evidence of a low concentration of HOD impurities. However, no spectral feature attributable to the bending mode of this species is observed in the region of interest (around λ = 6 μm). We then disregard any possible contribution of hydrogenated water species in the samples. The spectrometer employed for measuring the nonlinear IR spectra is based on a Micra-Legend Elite laser-amplifier system from Coherent Inc., Santa Clara, USA. The system produces a 1 kHz train of 35 fs (FWHM), 3 mJ pulses at 800 nm. Using 20% of this radiation, we generate by means of a home-built OPA25 and a TOPAS parametric generator (Light Conversion Ltd., Vilnius, Lithuania) mid-IR pulses of 1.5 and 2 μJ, respectively, with the central frequencies tunable between 1000 and 3000 cm1. The spectral width of both mid-IR pulses is about 200 cm1. We employ the TOPAS output and 6% of the OPA output as pump and probe pulses, respectively. Before the probe pulse reaches the sample, 50% of it is split-off and used as reference radiation. After the sample, both probe and reference pulses are spectrally dispersed in a spectrometer (TRIAX 180, HORIBA Jobin Yvon, Milano, Italy), and imaged separately on a 32 channel double array mercury cadmium telluride (MCT) detector (InfraRed Associates, Inc., Stuart, Florida, USA). The ratio of the probe to the reference spectra gives the third-order IR pumpprobe spectrum, which is recorded as a function of the pumpprobe time delay. As already mentioned, the full characterization of the guest molecule localization in the membrane requires determining the molecular conformation of the peptide, its relative orientation 5295
dx.doi.org/10.1021/jp109284z |J. Phys. Chem. B 2011, 115, 5294–5303
The Journal of Physical Chemistry B with respect to the neighboring phospholipids, and the depth of penetration inside the bilayer. The first two issues are addressed by one-color experiments: both pump and probe pulses are centered at 1680 cm1. Before being focused onto the sample, the pump beam traverses a tunable FabryPerot filter and a variable delay line. The FabryPerot element allows spectral narrowing of the pump pulse and its tuning across the spectral region of the carbonyl stretching modes.17 The nonlinear spectra were sampled in the delay time interval 01 ps. In our analysis we make use of the spectra taken at 500 fs delay, as under this experimental condition the cross-phase modulation contribution is absolutely negligible. The last issue was the subject of our previous experiment.14 Here we employ those results for a global comparison with the predictions of the MD simulation. The two-color experiment from which we extracted the information on the penetration depth of the guest peptide was described in detail in ref 14. Briefly, the broadband pump is centered at 2300 cm1, corresponding to the low frequency wing of the D2O stretching modes, while the broadband probe pulse is centered at 1680 cm1 and probes the CdO vibrations perturbed by the excitation of the water stretching modes. Molecular Modeling. MD simulations are carried out in the isothermalisobaric ensemble (T = 298 K and P = 0.1 MPa) using an orthorhombic simulation box with standard periodic boundary conditions. Temperature control is achieved using a Nose-Hoover thermostat,26,27 while constant pressure is imposed using a modification of the ParrinelloRahman Lagrangian.28 The PLPC molecules are modeled by the CHARMM force field,29,30 while the MrG molecules are modeled by the AMBER-99 force field variant developed by Hornak et al.31 The point-charge TIP3P model is used for water.32 LorentzBerthelot mixing rules account for Lennard-Jones interactions between different atom types. The phospholipid, dipeptide, and water molecules are completely flexible except for the covalent bonds involving hydrogen atoms, which are kept rigid. The equations of motion are integrated using the multiple time-step technique.33 The Ewald method with the smooth particle mesh algorithm3436 is applied to compute the electrostatic interactions. The simulated sample consists of 2877 water molecules and of two layers, each made of 63 PLPC molecules and one MrG molecule. The sample is prepared in two phases. In the first one we generate an equilibrated sample of a pure PLPC bilayer, starting from a previously equilibrated configuration of a hydrated PLPC bilayer made of 128 phospholipid molecules and 2460 water molecules (available from http://moose.bio.ucalgary.ca). The box size of this configuration is 79.666 54.478 55.065 Å. This sample corresponds to the final configuration of a MD simulation performed by Herrenbauer et al.37 In order to avoid possible artifacts due to unexpected peptide arrangement arising from periodic boundary conditions, the box dimension along the bilayer normal (from now on, the z-axis) is increased by =3.5 Å by adding 417 water molecules (to get a total of 2877 water molecules). The resulting hydrated PLPC bilayer is then equilibrated through a standard 5 ns MD simulation. Equilibration is verified by monitoring the potential energy, the simulation-box volume, and the stress tensor, which is found to be almost isotropic. In the second phase, we build up the initial configuration of the MrGPLPC sample by replacing two PLPC molecules, randomly chosen on the two leaflets of the pure PLPC sample, with MrG molecules. We point out that the replacement of PLPC molecules with MrG molecules does not introduce
ARTICLE
relevant frustration in the system because of the similarity between these two types of molecules. System equilibration is then achieved with a 2 ns isothermalisobaric MD simulation. With this procedure the PLPC molecules can be considered to be fully equilibrated, while the dipeptide molecules may still preserve memory of their initial configuration, due to their rugged energy landscape possessing several local minima where the peptide can be trapped during the simulation. Therefore, also considering the small number of MrG molecules, in order to accelerate the conformational sampling of the dipeptide molecules, we employ the parallel tempering technique,38 whose MD version is known as replica exchange molecular dynamics (REMD).39 In REMD simulations, several copies of the system, termed replicas, are simulated in parallel at different temperatures. At regular time intervals, exchanges of replicas at neighboring temperatures are attempted and accepted according to a Metropolis-like criterion. The advantage of this technique is that the acquisition of hot configurations that may result from the exchanges allows the peptide to escape from local energy minima, thereby enhancing the efficiency of the sampling at the target temperature. Actually, in our simulations we have adopted the so-called Hamiltonian-REMD method,40,41 a variant of the temperature-REMD described above. Hamiltonian-REMD differs from temperature-REMD by the fact that all replicas are simulated at the same temperature, each replica being associated with a different factor c introduced to scale the potential energy of the replicas. In this scheme, the potential energy of a replica i is therefore cEi. In our case, the scaling is applied only to the intramolecular potential energy of both MrG molecules, apart from the harmonic stretching and bending potential terms. We use 16 replicas swapping among the following sequence of states: c = 1.000, 0.898, 0.807, 0.725, 0.651, 0.585, 0.525, 0.472, 0.424, 0.381, 0.342, 0.307, 0.276, 0.248, 0.223, and 0.200. Replica exchanges are attempted every 8 fs. The replica configurations characterized by c = 1 have unscaled potential energy and therefore correspond to configurations sampled at 298 K. Instead, the replica configurations characterized by, e.g., c = 0.2 correspond to states where the intramolecular potential energy of the MrG molecules (excluding harmonic stretching and bending terms) have a virtual temperature of T/c = 298/0.2 = 1490 K. Since each replica can move through the c ensemble, the MrG molecules can easily overcome the internal energy barriers with significant improvement of the conformational sampling. Before performing the production REMD run, an equilibration of 1.68 ns is carried out. Atomic configurations are recorded every 0.4 ps during a production run of 5.4 ns. The statistical analysis is done on the c = 1 system configurations. Ramachandran plots calculated for c = 0.2 configurations (i.e., at high virtual temperature) show that the sampling involves a much broader conformational space for both MrG molecules. Interested readers can find the data in the Supporting Information. All simulations reported in this study are performed using the ORAC program.41 The quantum mechanical calculations are based on the density functional theory, as implemented in the Gaussian 03 program,42 using the B3LYP functional and the 6-31þþG(d,p) basis set. In these calculations, the myristoyl tail of the dipeptide is replaced with a methyl. We represent the molecular conformation of the peptide backbone in terms of the Ramachandran Φ and Ψ angles ranging from (180°, 180°) to (100°, 180°), with steps of 20° for both angles. The structures of these configurations are optimized by keeping the Φ and Ψ angles fixed. For all optimized configurations, we calculate the coupling constants, the angle 5296
dx.doi.org/10.1021/jp109284z |J. Phys. Chem. B 2011, 115, 5294–5303
The Journal of Physical Chemistry B
Figure 2. (A) FTIR spectrum of MrG in PLPC membrane. (B,C) Nonlinear 2D-IR spectra of MrG in PLPC membrane measured in parallel and perpendicular pumpprobe polarizations, respectively. (D) Zoomed view of the amide I resonances of panel B.
between the transition dipole moments, and the ratio between the IR intensities of the two amide I modes. These quantities are then given as functions of the Φ and Ψ angles within the Ramachandran space. To be specific, the values of the coupling constants between interacting CdO oscillators are obtained from ab initio calculations by reconstructing the Hessian Matrix for the normal modes in the subspace of the amide I modes.4346 Here we use the method described in ref 47, based on the Cartesian displacements of the normal modes (see Supporting Information).
’ RESULTS AND DISCUSSION Figure 2 represents the linear Fourier transform infrared (FTIR) and nonlinear 2D-IR spectra of the sample in the spectral range of the carbonyl stretching modes. The 2D-IR spectra, in parallel and perpendicular polarization, are recorded at 500 fs pumpprobe delay time. The band centered at 1734 cm1 is due to the vibration of the carbonyl groups of PLPC. The resonances of the 13C-labeled and of the native carbonyls of MrG are clearly separated in frequency (1587 cm1 and at 1632 cm1, respectively). The latter two bands are much narrower than the PLPC resonance, whose full width at half-maximum (FWHM) is about 40 cm1. We should notice that, for a similar phospholipid bilayer (dimyristoylphosphatidylcholine),11 the CdO stretching bandwidth was reduced by approximately 25% upon removal of the excitonic coupling by isotopic substitution. The 2D-IR spectrum shows a well-defined cross-peak between the two amide I bands of the dipeptide (pump centered at 1598 cm1, probe at 1632 cm1). In ref 14 we showed that the MrG molecule is embedded in the polar layer of the membrane. The resulting intermolecular hostguest coupling is confirmed by the appearance of cross-peaks between the two amide I transitions and the PLPC CdO stretching band (pump at 1598 and 1632 cm1, probe at 1740 cm1). The presence of
ARTICLE
rather intense intra- and intermolecular cross-peaks allows exploring structural properties of MrG anchored to the membrane. The close-to-Lorentzian lineshapes of the two amide I bands in the FTIR spectrum correspond to the essentially vertical alignment of the same two resonances in the 2D-IR spectrum of Figure 2. No detectable inhomogeneous character shows up in the linear and nonlinear spectra. Actually, in the analysis of the 2D-IR lineshape, one has to take into account the limitations due to the pump bandwidth (17 cm1) and to the actual spectral resolution of the detection system (5 cm1). On this basis, we cannot exclude that some small inhomogeneous contribution, on the order of very few wavenumbers, is present, resulting in a hardly detectable tilt toward the diagonal of the nodal line separating the negative and positive lobes of the 2D-IR resonances. An additional element to be considered is the relatively long pump pulse used in the experiment. The 17 cm1 bandwidth corresponds to a pulse duration of about 800 fs. With this low time resolution, a very fast (typically, < 300 fs) loss of memory of the vibrational excitation would lead to smearing out any trace of inhomogeneity, even at very short pumpprobe delay times. In Figure 2D, the lines represent the orientation of the nodal lines for the two amide I bands. Within the experimental sensitivity, the lines are essentially vertical. Under this respect, it is striking the difference with the spectral properties of, for instance, trialanine in water.15 In that case, in fact, a partial inhomogeneous character is observed, attributed to the copresence of different intramolecular conformational states of trialanine. The essentially homogeneous character of the amide I resonances of MrG suggests that this dipeptide is present in the phospholipid environment with one specific, largely prevalent conformation. The different behaviors of MrG in the membrane and of trialanine in water is at first sight contra-intuitive. In fact, the smaller size of the two glycine groups of MrG should in principle gives rise to a shallower backbone torsional potential and, consequently, to an easier access to different conformations. The very role of the anisotropic environment provided by phospholipid matrix and, possibly, the constraint due to the anchor tail, then come into play in “freezing” the conformational mobility of MrG. In the next sections, using the results of MD simulations and quantum mechanical calculations, we provide a substantial support for this interpretation, and we discuss in more details the mechanisms of the line narrowing of the MrG amide I resonances. Dipeptide Backbone Structure. The coupling constant between the amide I modes of the dipeptide, obtained from the relative intensities of the diagonal resonances and the crosspeaks, is 5 ( 1 cm1.48 From the cross-peak anisotropy, following the method proposed by Hochstrasser et al.,4850 we obtain the value of 43 (or 18043) ( 8° for the angle between the transition dipole moments of C13-labeled and native amide I modes. The confidence limits for these quantities were derived from a statistical analysis of a set of 48 nonlinear spectra obtained with different pump pulse frequencies and for four different samples (see Supporting Information). Then, by comparing the experimentally derived data with the calculated distributions in the Ramachandran Φ and Ψ space, we extract the compatible angular regions. The relative IR intensity of the two amide I bands is the additional property that we use to restrict the range of possible backbone conformations. According to our ab initio calculated IR spectra, the intensity ratio of the two transitions changes with the Φ and Ψ angles. In the experimental FTIR 5297
dx.doi.org/10.1021/jp109284z |J. Phys. Chem. B 2011, 115, 5294–5303
The Journal of Physical Chemistry B
ARTICLE
Figure 3. (AC) Coupling constant, angle between the transition dipole moments, and IR intensity ratio between 13C and 12C amide I modes of MrG calculated ab initio as a function of the Φ and Ψ angles. (DF) The colored areas are the Φ/Ψ regions compatible with the experimental values of the coupling constant, the angle between the transition dipole moments, and the IR intensity ratio, respectively.
spectrum, the intensity ratio of the 13C and 12C amide I bands is I13/I12 = 1.78. Note, however, that the calculations are done for a molecule in vacuum. Considering that the effect of the membrane environment on the IR intensities may be not negligible, we accept all the conformations that satisfy the condition I13/I12 > 1. Figure 3 gives the calculated Φ/Ψ maps (panels A, B, and C) and the regions corresponding to compatible values of the coupling constant, of the angle between transition dipole moments and of the intensity ratio (panels D, E, and F). Figure 4A represents the result of the combination of the three constraints on the structural maps. The acceptable region is very limited and corresponds to a distribution of structures of essentially unfolded character, with Φ and Ψ angles varying in the range [100°, 180°] and [190°, 120°], respectively. Figure 4B shows the conformational distribution of MrG in terms of the Φ and Ψ angles according to the REMD simulation. The Φ/Ψ distribution predicted by the simulation is definitely broader. We may single out two major structural configurations. The first one is essentially unfolded, with Φ and Ψ angles around 90 ( 20° and 180 ( 20°, respectively. It is compatible with the values of the dihedral angles obtained from the experiments (see Figure 4A). The second structural motif is close to a 310/R folded structure, with Φ and Ψ angles around 95 ( 20° and 10 ( 50°, respectively. The unfolded structure has the higher probability (64%), compared to the folded one (36%). Representative structures of the two main conformers are given in Figure 4. We already noticed that both linear and 2D-IR spectra in the amide I region are characterized by very low, practically undetectable, inhomogeneity. In order to verify whether the conformational distribution predicted by the REMD simulation is compatible with the experimental results, we calculate ab initio the IR spectra of
various folded and unfolded conformations of the dipeptide (using the Gaussian 03 program43). More precisely, we calculate the IR frequencies and intensities for a grid of points in the Φ/Ψ space (separated by 20°) according to the distribution of Figure 4B. Thus, each point in the Φ/Ψ grid contributes to the overall band profile with a Lorentzian line centered at the calculated frequency, with an intensity proportional to the product between the ab initio calculated intensity and the probability of the corresponding Φ/Ψ conformational state estimated in the REMD simulation. The spectral profiles of the folded and unfolded structures are then obtained as superposition of the relevant contributions. The calculated frequencies are scaled by 0.93 in order to fit the experimental spectrum. In Figure 5B,C, we report the reconstructed spectra for the folded and unfolded conformations of MrG, respectively. The comparison of the two calculated spectral profiles shows that, in the case of the 310/R folded structure, the two components have different intensities and, most importantly, the low frequency band is substantially broader. This latter feature is a consequence of the much higher value of the vibrational coupling between the two carbonyls in the folded geometry. This calculated spectrum is definitely inconsistent with the experimental profile. On the contrary, the spectrum calculated for the unfolded structures (Figure 5C) agrees well with the experiment. We may then conclude that the REMD simulation agrees with the experiments in predicting the unfolded backbone conformation as the most probable one. The presence of about 36% peptide molecules with folded backbone conformation appears instead to be overestimated by the simulation. The profiles of the linear IR and 2D-IR spectra indicate that the abundance of this conformation should be much less. The weak shoulders that appear at the bases of the two main bands are compatible with an abundance of other 5298
dx.doi.org/10.1021/jp109284z |J. Phys. Chem. B 2011, 115, 5294–5303
The Journal of Physical Chemistry B
ARTICLE
Figure 4. (A) Region of the Φ/Ψ space obtained from the intersection of the D, E, and F plots of Figure 3. (B) Probability distribution of the Φ and Ψ angles obtained from the REMD simulation. The two molecular structures on the right are representations of the two main backbone conformations (the green sphere represents the alkyl chain).
Figure 5. (A) FTIR spectrum of MrG in PLPC membrane (zoomed view of Figure 2A). (B) IR spectrum of the folded MrG conformations estimated with ab initio calculations (see Discussion). (C) IR spectrum of the unfolded MrG conformations estimated with ab initio calculations (see Discussion).
conformers on the order of 1015%. In evaluating this discrepancy, one has to consider that the actual conformational distribution is the result of the subtle competition of the intramolecular (basically torsional) potential and of the intermolecular peptidephospholipid interactions. In our simulation we chose to adopt the AMBER99 potential,31 developed and widely employed for peptides. In principle, it would be possible to remodel the intramolecular potential of MrG to account for the conformational distribution estimated from the experiments. However, the required computational effort would be well beyond the aim of this article. The calculated spectrum of Figure 5C also provides additional arguments for interpreting the unusually narrow line-shapes of the two amide I transitions reported in Figure 2 (and Figure 5A).
As already noticed, the positive and negative lobes of the 2D-IR bands are essentially vertical also at very early times, although the presence of a very small tilt toward the diagonal direction cannot be completely excluded, at least for the high frequency band. The bandwidth is about 14 cm1, in agreement with the value measured in the linear spectrum. On the other hand, Figure 5C shows that the calculated IR band-shapes, resulting from the frequency distribution of the unfolded conformation, are in qualitative agreement with the experiment, suggesting that the leading term determining the overall band shape is the (rather narrow) distribution of the backbone dihedral angles of the unfolded conformers, with a minor contribution from the environment. The bandwidths of the calculated spectrum of Figure 5C are, however, definitely broader (about 25 cm1) than the experimental ones. This discrepancy suggests that some motional narrowing effect has to be considered. In the previous section we noticed that, with the time resolution of our experiment, very fast (say 300 fs, or less) spectral diffusion would make a limited inhomogeneous contribution (about 10 cm1, in our case) practically undetectable. In the assumption that only the unfolded conformation of MrG is appreciably populated, two main processes can provide the required fast frequency modulation: the breaking and making of hydrogen bonds between the CdO groups and the water molecules present at the membrane interface, and the rapid small fluctuations of the dipeptide backbone around its equilibrium unfolded structure. The hydrogen bond lifetime obtained from a standard MD simulation (the details of this simulation are provided in the Supporting Information) is 2.3 ps for the 13CdO group of MrG (located deeper inside the polar layer of the membrane14) and 0.7 ps for the native carbonyl group of MrG, which is more exposed to hydration water. Therefore, it seems unlikely that the hydrogen bond dynamics can be responsible for the rapid frequency fluctuations suggested by the experimental results. It seems more reasonable to attribute such fast modulation to small amplitude fluctuations of the MrG backbone. This process was 5299
dx.doi.org/10.1021/jp109284z |J. Phys. Chem. B 2011, 115, 5294–5303
The Journal of Physical Chemistry B
ARTICLE
Figure 6. Weighted pair radial-angular distribution functions of the PLPC and MrG carbonyl groups (eq 1). (A) Function related to PLPC CdO and MrG 13CdO pairs. (B) Function related to PLPC CdO and MrG 12CdO pairs. The vertical dashed lines represent the values of the angles derived from the 2D-IR experiment; the corresponding errors are indicated by the horizontal bars.
found to contribute to the narrowing of the IR lineshape of trialanine amide I transition in water.15 Dipeptide Localization in the Membrane. (a). Angular Correlation. In a previous study we learned about the partially excitonic character of the delocalized carbonyl vibrations at the interface of the phospholipid membrane.11 In particular, as the dipeptide is present as an impurity in the membrane, and since the amide I modes are spectrally separated from the frequencies of the PLPC carbonyl modes, we may securely state that the spectral features of the cross-peaks between the amide I transitions and the PLPC CdO band can be interpreted in terms of intermolecular pairwise interactions. In other words, the cross-peaks measured at the PLPC carbonyl stretching frequency upon excitation of the amide I modes provide information on the average angle between the transition dipole moments of MrG and of the neighboring PLPC molecules. In the current experiment, from the measured anisotropy we obtain the values of 21 ( 14° and 33 ( 6° for the 13C and 12 C moieties, respectively (or equivalently 159° and 147°). To rationalize these values, we calculated the weighted radialangular distribution function related to MrG and PLPC carbonyl pairs. It is defined as * hðR, θÞ ¼
2 X
N X
+ βij δðR Rij Þδðθ θij Þ
ð1Þ
i¼1 j¼1 i ∈ MRG j ∈ PLPC
where the sum on i runs over a given type of carbonyl group of the two MrG molecules (either distal or vicinal to the myristoyl tail), while the sum on j runs over all PLPC carbonyl groups. The angular brackets indicate the average over all configurations of the REMD simulation. The quantity Rij is the distance between
the centers of the carbonyl groups, and θij is the angle formed by them. We calculate the coupling coefficient βij according to the transition dipole coupling model;51 the Torii and Tasumi method52 is used for determining the transition dipole moments of the amide I modes of MrG, while for the transition dipole moment of the PLPC carbonyl modes we follow the model proposed in refs 11 and 53. The calculated weighted radial-angular distribution functions are shown in Figure 6. For both MrG amide groups, the function is sharply peaked at the distance of 5 Å. In the case of the carbonyl next to the anchor tail (Figure 6A) also the angular distribution is quite narrow (between 20° and 50°). For the distal carbonyl instead, the angular distribution is definitely broader. The vertical dashed lines in the same figure correspond to the angles derived from the experiment. The calculated values are in general compatible with those obtained from the 2D-IR spectra, although the angular distribution calculated for the distal carbonyl is centered at a larger angle. In any case, it should be noted that the experimental error affecting the angular value of the distal carbonyl is much larger than that for the vicinal one (see error bars in Figure 6). (b). Depth of Penetration and Hydration of the Dipeptide. The depth of penetration of MrG into the PLPC membrane was the subject of our previous experimental work.14 We recall here briefly that the experiment consisted in exciting the water stretching modes and measuring the perturbation induced in the amide I transitions of MrG. The meter stick used for determining the vertical localization of MrG is provided by the density profile of the water molecules across the membrane, which is known from neutron diffraction measurements.54 In fact, the appearance of cross-peaks in the two-color 2D-IR spectra is a consequence of the hydrogen bonding between the CdO groups and the surrounding water molecules. The intensity ratio between the cross-peaks (Figure 7) involving the amide I resonances of MrG and that corresponding to the PLPC 5300
dx.doi.org/10.1021/jp109284z |J. Phys. Chem. B 2011, 115, 5294–5303
The Journal of Physical Chemistry B
Figure 7. Nonlinear differential IR spectrum obtained by exciting the stretching modes of D2O and probing the absorption in the region of the carbonyl stretching. Pumpprobe delay time is 500 fs.
Figure 8. (A) Atomic densities of water (green line) and PLPC carbonyl (black line) from neutron scattering data. The blue and red dotted lines are the altitudes of the 13C and 12C carbonyls of MrG, as deduced from the experiment of ref 14. (B) Distribution of all PLPC carbonyl oxygen atoms obtained from the REMD simulation (black line); distribution of the water oxygen atoms from the REMD simulation (green line); blue line shows the function calculated from REMD simulation by multiplying the distribution of the oxygen atoms of the 13 C labeled carbonyl of MrG with the distribution of water oxygen atoms; red line shows the same for the oxygen MrG native carbonyls. (C) Distribution of all PLPC carbonyl oxygen atoms obtained from the REMD simulation (black line); the blue and red curves represent the calculated distributions of the oxygens of the MRG carbonyls H-bonded to water molecules.55 The arrows indicate the average altitudes of the two distributions. For graphical reasons, in panels B and C the vertical axes are up-shifted to make the maxima of the experimental and calculated densities of the PLPC carbonyls coincident.
carbonyls (whose localization in the bilayer thickness is known from neutron scattering data54) is taken as a measure of the relative degree of hydration of the MrG carbonyls with respect to those of PLPC. From the known vertical profile of the water density, one can then derive the localization of the CdO groups of MrG. The results obtained following this procedure are summarized in Figure 8A. The two carbonyls of the dipeptide are found to lay, in average, above those of PLPC molecules: more precisely, the vertical distances from the average position of PLPC carbonyls was ∼2 Å for the carbonyl next to the MrG tail, and ∼4.4 Å for the distal carbonyl. Clearly, only the hydrated carbonyls can be detected by this method. The analysis of the REMD simulation described in the present paper provides a reliable test of this conclusion. In Figure 8B, the blue and red lines are obtained by multiplying the calculated distribution of the oxygen atoms of MrG by the density of water oxygen atoms as obtained from the REMD simulation. The black curve is the distribution of all carbonyl oxygens of the PLPC molecules. In agreement with the
ARTICLE
results of our two-color pumpprobe experiment, the REMD simulation predicts for the two amide groups an average localization above that of the PLPC carbonyls. The distances are slightly smaller (1.9 and 3.1 Å) than those deduced from the experiment. The method adopted in ref 14 and used here to calculate the curves in Figure 8B is based on the assumption that the water density depends only on the value of the z coordinate across the membrane thickness. However, from the simulation data we can exactly calculate the degree of hydration of the oxygen atoms of MrG and of the PLPC carbonyls in the sample. Figure 8C reports the distribution of the MrG oxygens hydrogen bonded to water molecules.55 Their average position is still above that of the PLPC carbonyls, but their distance from that reference altitude is even smaller: 0.8 and 2.3 Å. The obvious conclusion is that the assumption of the water density depending only on the z coordinate is not completely justified. In order to verify this point, we calculate the ratio between the water densities around MrG and PLPC molecules. In particular, we consider the water density in xy planes parallel to the membrane surface, as a function of the altitude in the perpendicular (z) direction. The origin of the z axis is placed at the average altitude of the carbonyl oxygens; the abscissa represents the distance in the xy plane from the carbonyl oxygens of MrG/ PLPC. The ratio of the two densities is given in Figure 9A. The figure shows that there is definitely more water near the MrG oxygens than near the PLPC carbonyl oxygens. The reason is clear from Figure 9B. This plot is calculated with the same procedure of Figure 9A, considering all atoms but excluding those belonging to water molecules. It is apparent that the plots of Figure 9A,B are somehow complementary. The increased density of water molecules corresponds to areas of the membrane (those above the guest MrG molecules) where the density of the polar groups of PLPC (namely phosphate and choline) is lower. The MrG molecule entering the membrane polar surface creates the pathway for a deeper penetration of water inside the membrane. This results in a different hydration of the CdO groups, the amide groups of MrG being more hydrated than the carbonyls of PLPC located at the same altitude. This is confirmed by the mean hydrogen bond number calculated for the MrG and PLPC carbonyls:55 the values are 0.87 for the former and 0.54 for the latter. If this picture corresponds to the structure of the real membrane, one should expect that the method adopted in ref 14 overestimates the average distance of the dipeptide from the center of the membrane. The structure and conformation of the anchor tail of MrG, in relation to that of PLPC are not directly observable in our experiments. It is, however, of some interest to analyze, under this respect, the sample obtained from the REMD simulation. To this purpose, we consider several structural properties, concerning both the anchor chain conformation and the mutual arrangement of the aliphatic tails of MrG and PLPC molecules. A more extensive discussion of these findings is given in the Supporting Information. Here we stress just two aspects: the penetration of the myristoyl anchor in the hydrophobic layer, and its conformation in terms of end-to-end distance. Figure 10A shows the distribution functions of the carbonyl-oxygen atoms and of the chain end-carbon atoms of MrG and PLPC along the z-axis of the membrane. It is noteworthy that the terminal carbon atom of the MrG anchor is centered near z = 0 and is very similar to that of the corresponding carbon atoms of the PLPC hydrocarbon tails. This implies that, on average, the anchor chain of MrG is parallel to the direction of the PLPC tails. This is confirmed by the plot of 5301
dx.doi.org/10.1021/jp109284z |J. Phys. Chem. B 2011, 115, 5294–5303
The Journal of Physical Chemistry B
ARTICLE
Figure 9. (A) Ratio between the densities of water oxygen atoms around MrG and PLPC carbonyl oxygens, respectively, as a function of the depth into the membrane of the MrG/PLPC carbonyl oxygen (distance along z axis) and of the distance in the membrane plane between water oxygen and MrG/ PLPC carbonyl oxygen (distance in the xy plane). (B) Same ratio as in panel A, but considering all nonwater atoms.
Figure 10. (A) Normalized distribution functions of the carbonyloxygen atoms (carbonyl-O) and of the chain end-carbon atoms (End-C) of MrG and PLPC along the z-axis of the membrane. (B) Normalized distribution functions of the distance between the first (next to the carbonyl group) and the last carbon atoms of the MrG, sn1-PLPC and sn2-PLPC hydrocarbon tails (see Figure 1 for sn labeling notation).
Figure 10B, where we show the distribution functions of the distances between the first (next to the carbonyl group) and the last carbon atoms of the hydrocarbon tails of MrG and PLPC molecules (end-to-end distance). This is a simple criterion for quantifying the degree of folding of the long aliphatic chains. The figure shows that only the linoleic (sn2) chain of PLPC, whose distribution function shows a not negligible wing reaching values lower than 4 Å, is present in folded conformations. In fact, it is known that unsaturated hydrocarbon chains of lipidic bilayers can be found in folded geometries. On the contrary, the myristoyl chain of MrG and the palmitic (sn1) tail of PLPC have almost identical length distributions (apart from the obvious difference due to the different number of constituent carbon atoms), corresponding to prevalent unfolded structures.
’ CONCLUSIONS Understanding the structural properties of the association between a small peptide molecule and a phospholipid membrane is a complex task, which requires the coordinated use of a number of experimental and theoretical tools. Here we show that, by means of nonlinear infrared spectroscopy methods employing isotopic labeling, it is possible to learn about the intramolecular conformation of the dipeptide backbone, its relative orientation within the membrane polar layer, and the depth of penetretation inside the bilayer. The results of the computer simulations are in fair agreement with the structural properties deduced from the
spectroscopic investigation, thus substantiating and reinforcing the interpretation of the experimental data. The MrG dipeptide is found to be very strongly associated to the membrane, with its anchoring chain deeply inserted into its hydrophobic layer. The distribution of the angle formed by the peptide backbone and the normal to the surface is centered at about 60°. The largely prevalent conformation of the molecule corresponds to an unfolded structure, with Φ and Ψ dihedral angles centered at 100° and 180°, respectively. The amide groups of MrG lay slightly above the plane of the phospholipid carbonyls, while the distribution of the angles between the CdO groups of the peptide and those of the neighbor phospholipids is peaked at rather small angles (typically, between 20 and 50 degrees). The simulation also shows that the water distribution within the polar layer of the membrane is modified in the region occupied by the guest peptide molecules. The reduced density of nonwater atoms in the region above the head of the peptide (mainly determined by the absence of phosphate and choline) allows an easier access of the water molecules to the inner part of the layer. The successful application of the diversified and coordinated investigation methods adopted here, including nonlinear IR spectroscopic techniques and molecular modeling, represents a strong indication that similar approaches can be used to tackle structural problems of higher complexity, such as those involving polypeptides of medical and pharmacological interest.
’ ASSOCIATED CONTENT
bS
Supporting Information. (a) Capability of REMD in exploring the conformational space of the MrG dipeptide. (b) Hessian matrix reconstruction. (c) Structural analysis from REMD simulation. (d) H-bond dynamics. (e) Angle between the amide I transition dipole moments and coupling between the amide I modes of MrG. This material is available free of charge via the Internet at http://pubs.acs.org.
’ ACKNOWLEDGMENT The authors are grateful to Alessandro Barducci for setting up the PLPC force field in the ORAC program. The calculations were performed on a computer cluster at LENS with the technical help of Gianfranco Lauria (LENS and Department of Physics of the University of Firenze, Italy). This work has been supported by the European Union contract RII3-CT-2003-506350. 5302
dx.doi.org/10.1021/jp109284z |J. Phys. Chem. B 2011, 115, 5294–5303
The Journal of Physical Chemistry B
’ REFERENCES (1) Anderson, H. J.; Coleman, J. E.; Andersen, R. J.; Roberge, M. Cancer Chemother. Pharmacol. 1997, 39, 223. (2) Bai, R.; Durso, N. A.; Sackett, D. L.; Hamel, E. Biochemistry 1999, 38, 14302. (3) Loganzo, F.; Discafani, C. M.; Annable, T.; Beyer, C.; Musto, S.; Hari, M.; Tan, X.; Hardy, C.; Hernandez, R.; Baxter, M.; Singanallore, T.; Khafizova, G.; Poruchynsky, M. S.; Fojo, T.; Nieman, J. A.; AyralKaloustian, S.; Zask, A.; Andersen, R. J.; Greenberger., L. M. Cancer Res. 2003, 63, 1838. (4) Piekarz, R. L.; Robey, Rob.; Sandor, V.; Bakke, S.; Wilson, W. H.; Dahmoush, L.; Kingma, D. M.; Turner, M. L.; Altemus, R.; Bates, S. E. Blood 2001, 98, 2865. (5) Hamm, P.; Lim, M.; Hochstrasser., R. M. J. Phys. Chem. B 1998, 102, 6123. (6) Asplund, M. C.; Zanni, M. T.; Hochstrasser, R. M. Proc. Natl. Acad. Sci. U.S.A. 2000, 97, 8219. (7) Golonzka, O.; Khalil, M.; Demirdoven, N.; Tokmakoff, A. Phys. Rev. Lett. 2001, 86, 2154. (8) Pshenichnikov, M.; de Boeij, W.; Wiersma, D. Phys. Rev. Lett. 1996, 76, 4701. (9) de Boeij, W.; Pshenichnikov, M.; Wiersma, D. Chem. Phys. Lett. 1995, 238, 1. (10) Brixner, T.; Stenger, J.; Vaswani, H. M.; Cho, M.; Blankenship, R. E.; Fleming, G. R. Nature 2005, 434, 625. (11) Volkov, V. V.; Chelli, R.; Zhuang, W.; Nuti, F.; Takaoka, Y.; Papini, A. M.; Mukamel, S.; Righini, R. Proc. Natl. Acad. Sci. U.S.A. 2007, 104, 15323. (12) Mukherjee, P.; Krummel, A. T.; Fulmer, E. C.; Kass, I.; Arkin, I. T.; Zanni, M. T. J. Chem. Phys. 2004, 120, 10215. (13) Volkov, V.; Hamm, P. Biophys. J. 2004, 87, 4213. (14) Volkov, V.; Righini, R. J. Phys. Chem. B 2009, 113, 16246. (15) Woutersen, S.; Pfister, R.; Hamm, P.; Mu, Y.; Kosov, D. S.; Stock, G. J. Chem. Phys. 2002, 117, 6833. (16) la Cour Jansen, T.; Zhuang, W.; Mukamel, S. J. Chem. Phys. 2004, 121, 10577. (17) Woutersen, S.; Hamm, P. J. Phys. Chem. B 2000, 104, 11316. (18) Ding, F.; Zanni, M. T. Chem. Phys. 2007, 341, 95. (19) James, G.; Olson, E. N. Biochemistry 1990, 29, 2623. (20) Farazi, T. A.; Waksman, G.; Gordon, J. I. J. Biol. Chem. 2001, 276, 39501. (21) Luckey, M. Membrane Structural Biology with Biochemical and Biophysical Foundations; Cambridge University Press: New York, 2008. (22) Efremov, R. G.; Nolde, D. E.; Volynsky, P. E.; Chernyavsky, A. A.; Dubovskii, P. V.; Arseniev, A. S. FEBS Lett. 1999, 462, 205. (23) Gordon, L, M.; Mobley, P. W.; Pilpa, R.; Sherman, M. A.; Waring, A. J. Biochim. Biophys. Acta 2002, 1559, 96. (24) IDM Pharma Submits New Drug Application to the FDA for Junovan (mifamurtide) in the Treatment of Osteosarcoma, New Drug Applications Archive. http://www.drugs.com/nda/junovan_061026. html. IRVINE, Calif., October 26, 2006. (25) Kaindl, R. A.; Wurm, M.; Reimann, K.; Hamm, P.; Weiner, A. M.; Woerner, M. J. Opt. Soc. Am. B 2000, 17, 2086. (26) Hoover, W. G. Phys. Rev. A 1985, 31, 1695. (27) Hoover, W. G. Phys. Rev. A 1986, 34, 2499. (28) Marchi, M.; Procacci, P. J. Chem. Phys. 1998, 109, 5194. (29) Feller, S. E.; MacKerell, A. D. J. Phys. Chem. B 2000, 104, 7510. (30) Feller, S. E.; Gawrisch, G.; MacKerell, A. D. J. Am. Chem. Soc. 2002, 124, 318. (31) Hornak, V.; Abel, R.; Okur, A.; Strockbine, B.; Roitberg, A.; Simmerling, C. Proteins 2006, 65, 712. (32) Jorgensen, W. L.; Chandrasekhar, J.; Madura, J. D.; Impey, R. W.; Klein, M. L. J. Chem. Phys. 1983, 79, 926. (33) Tuckerman, M.; Berne, B. J.; Martyna, G. J. J. Chem. Phys. 1992, 97, 1990. (34) Hockney, R. W. Computer Simulation Using Particles; McGrawHill: New York, 1989.
ARTICLE
(35) Darden, T.; York, D.; Pedersen, L. G. J. Chem. Phys. 1993, 98, 10089. (36) Essmann, U.; Perera, L.; Berkowitz, M. L.; Darden, T.; Lee, H.; Pedersen, L. G. J. Chem. Phys. 1995, 101, 8577. (37) Herrenbauer, M.; Tieleman, D. P.; Posten, C. Computer Applications in Biotechnology 2001, 8, 317. (38) Hukushima, K.; Nemoto, K. J. Phys. Soc. Jpn. 1996, 65, 1604. (39) Sugita, Y.; Okamoto, Y. Chem. Phys. Lett. 1999, 314, 141. (40) Fukunishi, H.; Watanabe, O.; Takada, S. J. Chem. Phys. 2002, 116, 9058. (41) Marsili, S.; Signorini, G. F.; Chelli, R.; Marchi, M.; Procacci, P. J. Comput. Chem. 2010, 31, 1106. (42) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Montgomery, Jr. J. A.; Vreven, T.; Kudin, K. N.; Burant, J. C. et al. Gaussian 03, revision C.02; Gaussian, Inc.: Wallingford, CT, 2004. (43) Torii, H.; Tasumi, M. J. Raman Spectrosc. 1998, 29, 81. (44) Hamm, P.; Woutersen, S. Bull. Chem. Soc. Jpn. 2002, 75, 985. (45) Ham, S.; Cho, M. J. Chem. Phys. 2003, 118, 6915. (46) Eker, F.; Cao, X.; Nafie, L.; Huang, Q.; Schweitzer-Stenner, R. J. Phys. Chem. B 2003, 107, 358. (47) Choi, J.-Ho; Ham, S.; Cho, M. J. Phys. Chem. B 2003, 107, 9132. (48) Hamm, P.; Lim, M.; Degrado, W. F.; Hochstrasser, R. M. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 2036. (49) Hochstrasser, R. M. Chem. Phys. 2001, 266, 273. (50) Zanni, M. T.; Ge, N.-H.; Kim, Y. S.; Hochstrasser, R. M. Proc. Natl. Acad. Sci. U.S.A. 2001, 98, 11265. (51) Krimm, S.; Bandekar, J. Adv. Protein Chem. 1986, 38, 181. (52) Torii, H.; Tasumi, M. J. Chem. Phys. 1992, 96, 3379. (53) Torii, H.; Tasumi, M. J. Chem. Phys. 1993, 99, 8459. (54) Jacobs, R. E.; White, S. H. Biochemistry 1989, 28, 3421. (55) In our analysis of the MD data, the carbonyl oxygen and the water hydrogen (deuterium) atoms are assumed to be hydrogen bonded if their distance is less than 2.4 Å.
5303
dx.doi.org/10.1021/jp109284z |J. Phys. Chem. B 2011, 115, 5294–5303
