Positioning and Stabilization of Dynorphin ... - ACS Publications

Dec 5, 2001 - Positioning and Stabilization of Dynorphin Peptides in Membrane Bilayers: the Mechanistic Role of Aromatic and Basic Residues Revealed ...
0 downloads 0 Views 672KB Size
J. Phys. Chem. B 2002, 106, 209-218

209

Positioning and Stabilization of Dynorphin Peptides in Membrane Bilayers: the Mechanistic Role of Aromatic and Basic Residues Revealed from Comparative MD Simulations Ramasubbu Sankararamakrishnan and Harel Weinstein* Department of Physiology and Biophysics, Mount Sinai School of Medicine, New York UniVersity, New York, New York 10029 ReceiVed: June 6, 2001; In Final Form: October 11, 2001

There is still a significant gap in our understanding of the physicochemical properties of membrane systems with embedded peptides. This limits the ability to understand key biological mechanisms of systems in which the effects of peptides and small proteins are determined by their environment. In previous studies of the 17-residue endogenous opioid peptide dynorphin A [(Dyn A(1-17)], a selective ligand of κ-opioid receptors that is important as an analgesic, we have shown from molecular dynamics (MD) simulations in dimyristoylphosphatidylcholine (DMPC) that the peptide maintains a defined secondary structure and a characteristic position within the membrane bilayers (Sankararamakrishnan and Weinstein, Biophys. J. 2000, 79, 2331-2344). Because some fragments of Dyn A(1-17) have cognate pharmacological properties, we studied the behavior of the smaller opioid [Dyn A(1-13)] and the nonopioid fragment of dynorphin [des-Tyr dynorphin; Dyn A(2-17)] in DMPC bilayers with MD simulations for periods of 9 and 6.5 ns, respectively. Dyn A(1-13) was found to behave similarly to the full length peptide Dyn A(1-17) in the bilayer. Its N-terminal helical segment that was initially oriented perpendicular to the membrane plane remained imbedded within the bilayers throughout the simulations and adopted a tilt angle of ∼35° with respect to the bilayer normal. Analysis of the peptide-membrane interactions reveals the key role of arginine and lysine residues which, as in Dyn A(1-17), are organized structurally in “snorkel-model” type bonding and contributed significantly toward peptide-DMPC and peptide-water interaction energies. In contrast, the absence of tyrosine in Dyn A(2-17) is found to yield (a) deeper membrane penetration of the helical segment and (b) a smaller extent of water penetration into the bilayers. The positioning of both Dyn A(1-17) and Dyn A(1-13) is determined by interactions of the aromatic residues tyrosine and phenylalanine, which are shown to prefer different components of the lipid bilayers. The nature of the dramatic differences in the mode of interactions with the membrane observed for the des-Tyr congener suggests a general mechanistic role for Tyr and Phe residues in determining both the physicochemical and the biological properties of peptides in the membrane environment. The importance of this role is illustrated by Tyr-1. We show that this residue is keeping the pharmacologically active Dyn A(1-17) peptide close to the membrane-water interface, a position that may be a determinant factor in the binding mechanism of the peptide with opioid receptors.

Introduction The endogenous peptide dynorphin A [Dyn A(1-17)] (HTyr-Gly-Gly-Phe-Leu-Arg-Arg-Ile-Arg-Pro-Lys-Leu-Lys-TrpAsp-Asn-Gln-OH1) binds selectively to the κ-opioid receptor and its actions on opioid receptors are antagonized by naloxone.2,3 Shorter fragments of Dyn A(1-17) also exhibit specific affinity and pharmacological potency at κ-opioid receptors.4,5 The peptide incorporating the first 13 residues [Dyn A(1-13)] of the naturally occurring Dyn A(1-17) has practically the same pharmacological profile as its parent peptide.2 As a result, Dyn A(1-13) is being used in many experimental studies in place of Dyn A(1-17). The des-Tyr dynorphin does not bind to the opioid receptors but it shares with dynorphin A a variety of biological functions that are nonopioid in nature.6-11 Experimental results from these studies suggest that these nonopioid actions are due to the binding of dynorphin congeners with the * To whom correspondence should be addressed. Department of Physiology and Biophysics, Box 1218, Mount Sinai School of Medicine, One Gustave Levy Place, New York, NY 10029. E-mail: hweinstein@ inka.mssm.edu. Fax: (212) 860 3369.

NMDA receptors, or other nonopioid receptors, or even a new yet-to-be-discovered dynorphin-specific receptor.7,12 Dynorphin’s selectivity for κ-receptors and the role of the first tyrosine residue in binding the opioid receptors appear to be directly dependent on the physicochemical properties that determine the behavior of the peptide in membrane bilayers. Studies of the explicit atomic interactions between the peptide and the receptors are likely to require an understanding of the manner in which the peptide hormones behave both in the aqueous medium and in the membrane. Schwyzer13,14 had proposed that the peptide hormones will first acquire an active conformation by interaction with the membrane. Experimental support for this proposal came from the NMR studies on cholecystokinin (CCK) peptide and its lipophilic adduct.15 Both the CCK peptide and the lipid-bound derivative showed comparable binding affinities to CCK receptors. More recently, NMR studies on pituitary adenylate cyclase activating peptide (PACAP) showed that the conformation bound to the receptor (in the class of GPCRs: G protein coupled receptors) is similar to that in micelles, with differences observable mainly in the

10.1021/jp012174o CCC: $22.00 © 2002 American Chemical Society Published on Web 12/05/2001

210 J. Phys. Chem. B, Vol. 106, No. 1, 2002 first few N-terminal residues.16 Studies on agonists and antagonists of mammalian tachykinin NK1 receptor,17 Neuropeptide Y,18 and analogues of bradykinin19 also support the idea that membranes play a vital role in inducing and stabilizing the structures of peptide hormones. Notably, however, the interactions with the membrane seem to have a degree of specificity, and the binding characteristics of κ-opioid ligands with the membrane medium are believed to be significantly different from those that are specific to δ and µ receptors.14 The structures and physicochemical properties of κ-selective dynorphin peptides, Dyn A(1-17) and Dyn A(1-13), have been studied in different solutions.20-22 While Dyn A(1-13) was found to be in extended conformation in aqueous solution,20 it assumed a partial helical structure in neutral lecithin membranes.21 Recently, NMR studies of Dyn A(1-17) in dodecylphosphocholine (DPC) micelles23 have shown that residues 3-9 adopt an R-helical conformation. Schwyzer suggested that when bound to the membrane, the more hydrophobic N-terminal helical segment of dynorphin would be oriented perpendicular to the membrane surface contacting the hydrophobic membrane layers, while the hydrophilic C-terminal segment would be in contact with the aqueous phase.24 To investigate the molecular details of the interactions of these peptides with membranes, we have recently carried out two multi-nanosecond molecular dynamics (MD) simulations on Dyn A(1-17) in dimyristoylphosphatidylcholine (DMPC) bilayers.25 The results showed that the peptides adopted defined structures and positions in the DMPC bilayers, independent of the starting geometries for the simulations. Thus, after >2 ns simulations, the structures from two different starting points converged to the same final structure. The N-terminal helical segment, which was initially oriented parallel to the membrane normal, assumed a tilted orientation at an angle of ∼50° with respect to the bilayer normal. Specific interactions of Dyn A(1-17) residues with phospholipids and waters were found to be responsible for stabilization and orientation of Dyn A(1-17) within the DMPC bilayer. In particular, the Tyr-1 residue played a key role by its preference to be close to the lipid headgroups, while the Phe-4 residue consistently pointed toward the center of the bilayer. In the study presented here, we have focused on the structural and mechanistic role of the Tyr-1 residue that is important for determining the physicochemical properties of the membraneimbedded peptide. The opioid and nonopioid fragments of Dyn A(1-17), namely Dyn A(1-13) and Dyn A(2-17), were simulated in DMPC bilayers using the technique of MD for a period of 6.5-9 ns. To distinguish the determinant elements in the structural effect of specific interactions with components of the membrane environment, we carried out for comparison purposes, a 4 ns MD simulation of Dyn A(1-17) in water. The results show that the behavior of Dyn A(1-13) is similar to Dyn A(1-17) within the DMPC bilayers, but that the des-Tyr congener, Dyn A(2-17), behaves differently. The physicochemical underpinnings of this difference in behavior are of interest because they are likely to be generalizable to peptides and membrane-spanning domains containing similarly positioned aromatic residues. Moreover, they seem to correlate with major differences in biological properties of the peptides. Analysis of the interactions of individual residues with lipids and waters, as well as solvation profiles and water penetration served in this study to determine the specific role of the key residues in the different modes of stabilization of dynorphin and its derivatives (especially des-Tyr1) in the phospholipid bilayer.

Sankararamakrishnan and Weinstein Materials and Methods Dynorphin Peptide Simulations in DMPC Bilayers. To enable extensive comparisons, we have used the same protocols for the construction of initial structures of dynorphin/lipid complexes and the simulation protocol, as described in our recent paper.25 The methods are reviewed below only briefly, and any differences from the previous protocol are explained in detail. The program CHARMM26 was used in all the simulations with an all-atom PARAM 22 force filed27 that included force fields for phospholipids28 and TIP3P water.29 The simulations used the NVE (constant number of atoms, volume and energy) microcanonical ensemble and were carried out at 330 K, well above the gel-liquid-phase transition of DMPC.30 Although constant pressure conditions have been recommended for lipid simulations,31 it should be pointed out that simulations on gramicidin channel using NPγT (constant number of atoms, pressure, nonzero surface tension, and temperature)32 and NVE33 (constant number of atoms, volume, and energy) ensembles resulted in similar properties of peptide and boundary lipids. These two simulations were performed at two different temperatures (305 and 340 K). The force field parameters and the nonbonded options in the present simulations were the same as in our previous studies.25 a. Initial Structures. The initial structures of dynorphin fragments were generated from the internal parameters obtained from the NMR studies of Dyn A(1-17) in DPC micelles.23 The segment containing residues 1-9 (1- in des-Tyr peptide) was built as an R-helical structure, and the residues 10 -13 (9-12 in des-Tyr peptide) were in random conformation. In Dyn A(217), residues 13-16 assumed a type I β-turn. The initial peptide and hydrated bilayer system was constructed using a protocol developed by Roux and his collaborators.33-35 The crosssectional area for DMPC was taken as 63.1 Å2. This value is reasonable since the estimated cross-sectional area for DMPC from experiments at 50 °C is 63.3 Å2.36 The periodic boundary system consists of a rectangular box in which X and Y dimensions are the same (X ) 53.3 Å, Y ) 53.3 Å). The center of the bilayer is at Z ) 0 Å and the membrane normal is oriented along the Z-axis. The peptide was placed in the upper layer (Z ) +ve) with 41 lipids, while the bottom layer consisted of 45 lipids. The difference in the number of lipids in the layers approximately corresponds to the cross-sectional area of the dynorphin peptides. DMPC molecules were randomly chosen from a library of 2000 preequilibrated37 and prehydrated38 DMPC lipids and their headgroups were positioned within Z ) (17 Å. Bad contacts between individual molecules (lipids, peptide, and waters) were removed by rigid body rotations and translations. The remaining bulk solvent was constructed as described previously.25 In Dyn A(2-17) simulations, the primary box along the Z-axis (Z ) 90 Å) was somewhat larger so that the longer C-terminal segment can be accommodated within the primary box throughout the simulations. The number of water molecules was high on both sides of the bilayer (62 waters/lipid), similar to our previous simulations on Dyn A(117).25 For Dyn A(1-13), the number of water molecules was taken close to the experimental value36 (∼30 waters/lipid) and the Z-axis was reduced to ∼61.5 Å. Thus the systems contained ∼26,400 atoms (peptide + 86 lipids + ∼5300 waters) for Dyn A(2-17) simulation and ∼18 300 atoms (peptide + 86 lipids + ∼2600 waters) for Dyn A(1-13) simulation. In both systems, the N-terminal helical segment was initially oriented perpendicular to the membrane plane within the bilayer, and the center of the CR atoms of residues 1-10 (residues 1-9 for des-Tyr peptide) was initially placed at Z ) 10 Å. The C-terminal residues were approximately parallel to the membrane plane.

Dynorphin Peptides in Membrane Bilayers

J. Phys. Chem. B, Vol. 106, No. 1, 2002 211

Figure 1. Initial position of Dyn A(2-17) (left) and Dyn A(1-13) (right) within DMPC bilayers after minimization. The following color code is used: lipid nitrogen, dark blue; phosphorus, yellow; carbonyls, red; waters, light blue; peptide, pink. The N-terminal helical segments are oriented perpendicular to the membrane plane. The water layer in Dyn A(2-17) is larger (62 waters/lipid) to accommodate the longer C-terminal segment within the primary box throughout the simulations. The number of water molecules in Dyn A(1-13) simulation is close to the experimental value.36

The systems were refined by energy minimization, and periodic boundary conditions were applied in all directions. The energyminimized structures for Dyn A(2-17) and Dyn A(1-13) are shown in Figure 1. b. Simulation Details. The systems were first coupled to a heat bath at 330 K, and Langevin dynamics simulation of 0.1 ns was carried out. In the initial stage, a planar harmonic restraint was applied at the center of mass of the lipid atoms to maintain the planarity of the membrane. A cylindrical harmonic restraint was applied on the center of mass of the R-helical segment to maintain R-helical conformation of the N-terminal helical segments. During the equilibration, the force constants of both restraints were gradually reduced. The equilibration period was 1.5 ns for Dyn A(2-17) and 1 ns for Dyn A(1-13). No restraints were applied during the last 400 ps of equilibration and in the subsequent production runs. Production trajectories consisted of 5 ns [Dyn A(2-17)] and 8 ns [Dyn A(1-13)] MD simulations. A time step of 0.002 ps was used in all the simulations. The coordinates were saved every 1 ps for analysis. The total energy, its decomposition into kinetic and potential energies, and the temperatures of the systems were monitored throughout the simulations. Analysis of the MD trajectories of the potential energies show that during the production runs, the values of the energies increased by about 100-200 kcal/mol [from ∼-63 190 to -63 270 kcal/mol for Dyn A(2-17); from ∼-37 160 to ∼-37 250 kcal/mol for Dyn A(2-13)]. This is an increase of about 0.5% from the energy value at the beginning of the production runs. The temperature of the system increased to 335 K in Dyn A(2-17) and 333 K in Dyn A(1-13) simulations. Dynorphin A(1-17) Simulation in Water. Dyn A(1-17) was simulated at 300 K in a sphere of waters using our recently published protocol.39 In this protocol, Dyn A(1-17) was imbedded in a water sphere of 35 Å radius. Before the MD simulation started, water molecules within 2.8 Å from the peptide were deleted and the peptide-water complex was energy

minimized with ∼17 800 atoms. During the simulation, a spherical harmonic restraint was applied on water molecules that are more than 30 Å away from the center of the sphere. A small force constant of 0.007 (kcal/mol)/Å2 was applied. This force constant has been shown to eliminate the potential artifacts of the surface effects and to prevent the water molecules from evaporating.39 Water structural and dynamical properties are maintained at the reference temperature in this protocol. During 50 ps equilibration, positional restraints were applied on the CR atoms of the peptide. In the subsequent 1.4 ns simulation, the restraints were gradually removed. A further 3 ns simulation was carried out as a production trajectory. A harmonic restraint on the center of mass of the peptide was applied to keep the peptide close to the center of the sphere during the entire period of simulation. Coordinates were saved every 10 ps for the analysis. Results The structures at the end of the production runs for the peptides in the bilayer are shown in Figure 2. Both the energies and the temperatures remained constant during the last 2 ns [Dyn A(2-17)] and 4 ns [Dyn A(1-13)] of the production runs. As in our previous simulations on Dyn A(1-17),25 the N-terminal helical segment remains imbedded within the bilayers throughout the simulations of both peptides. Visual examination shows that the helical segment of Dyn A(1-13) adopts an orientation similar to the parent peptide Dyn A(1-17).25 In the case of Dyn A(2-17), the helix penetrates deeper into the membrane bilayer. N-Terminal Helix Structure and Orientation. The average values of backbone dihedral angles φ and ψ are shown in Figure 3a for residues 1-10 in Dyn A(2-17) and Dyn A(1-13). Residues 4-9 are shown to be in R-helical conformation throughout the simulations in Dyn A(1-13). Although this is largely true for Dyn A(2-17) as well, there is a slight distortion at residue 7. The orientations of the helical segment were

212 J. Phys. Chem. B, Vol. 106, No. 1, 2002

Sankararamakrishnan and Weinstein

Figure 2. Positions of the peptides Dyn A(2-17) (left) and Dyn A(1-13) (right) in the DMPC bilayers at the end of 5 and 8 ns production runs. Water molecules are omitted for clarity. The helical segments remain imbedded within the bilayers in both simulations. The helical segment of Dyn A(1-13) adopts an orientation similar to that of Dyn A(1-17).25 The helix of Dyn A(2-17) penetrates 7 Å deeper inside the bilayer compared to Dyn A(1-13).

Figure 3. (A) Average backbone dihedral angles φ (b) and ψ (9) calculated for the entire production runs for the N-terminal helical regions of Dyn A(1-13) (blue) and Dyn A(2-17) (red). To facilitate the comparison, the numbering of Dyn A(2-17) begins from 2 in this and subsequent figures. (B) Normalized MD trajectories of the center of mass location of the N-terminal helical segments along the bilayer normal. The dotted curves represent the average positions of midpoints of P and N atoms of the top layer. In addition to Dyn A(1-13) (blue) and Dyn A(2-17) (red), data obtained from the two simulations of Dyn A(1-17) (black)25 are also shown. The analysis is shown for the last 1 ns of the production runs from each simulation.

determined by the angle between the helix axis and the bilayer normal (Z-axis). The depths of the helix within the bilayers were

monitored from the MD trajectories of the center of mass location of the helix segments along the bilayer (Figure 3b). For comparison, trajectories from the simulations25 of Dyn A(117) are also shown. Analysis of the last 1 ns production run showed that the position of Dyn A(1-13) remained close to that of Dyn A(1-17). However, the helix in Dyn A(2-17) has penetrated deeper by about 7 Å in comparison to Dyn A(113). While there is a significant difference in the helix penetration within the membrane, the orientation of the helix segments with respect to the bilayer normal is almost the same in the two simulations. The helix segments make an angle of 30-35° with the bilayer normal and this value was close to 50° in Dyn A(1-17) simulations. Because analysis of helix depth and orientation within the bilayers indicates that the absence of Tyr-1 residue has a significant impact on the way des-Tyr dynorphin interacts with the membrane, we examined the role of individual residues in the membrane-insertion mechanism of dynorphin peptides. Density and solvation profiles of individual residues, interaction energies of the peptides with lipids/water, and the penetration of water molecules within the bilayers are presented below for the last 1-2.5 ns of the production runs. To facilitate comparison, the numbering of Dyn A(2-17) in the subsequent sections begins at 2. Density Profiles. Average density profiles of individual residues of Dyn A(1-13) and Dyn A(2-17) along the bilayer normal (Z-axis) are shown in Figure 4. A comparison of the profiles shows that the residues in Dyn A(2-17) have moved deeper inside the bilayer. The density profile of Phe-4 is wider than that of Tyr-1 in Dyn A(1-13) and the residues overlap in the region of 5-10 Å along the bilayer normal. The distribution of Phe-4 toward the bilayer center and that of Tyr-1 upward, beyond the next two consecutive glycine residues in the positive Z-direction, reveal that Phe-4 and Tyr-1 are pointing in the opposite directions along the bilayer normal. The same tendency of these aromatic residues was observed in the simulations of Dyn A(1-17).25 Because of its deeper penetration, the profile of the Phe residue in Dyn A(2-17) is spread beyond the bilayer center Z ) 0 Å, indicating possible interactions of Phe with the lipid acyl chains from the other-half of the bilayer. In Dyn A(2-17) the absence of the tyrosine residue in the first position has shifted the density profiles of all residues toward the interior of the bilayer. The distribution of arginine and lysine residues falls in the same region as the lipid headgroup components choline, phosphate, and carbonyls (data not shown). As for DynA(1-17),25 strong interactions of basic residues with the

Dynorphin Peptides in Membrane Bilayers

J. Phys. Chem. B, Vol. 106, No. 1, 2002 213

Figure 4. Average density profiles of individual amino acid residues (first 10 residues only) of Dyn A(1-13) (top) and Dyn A(2-17) (bottom). The profiles were computed over the structures saved during the last 1 ns of production runs for both simulations. Density profiles of Tyr-1 and Phe-4 are plotted in green and red, respectively. The three arginine residues are in blue and the hydrophobic residues are in black.

lipid headgroup and water are expected to play important roles in keeping the N-terminal helical segment within the bilayers. The Relation of Physicochemical Properties to Solvation and Positioning. a. SolVation of Side Chains. The solvation pattern of individual dynorphin residues was obtained by counting the number of nearest neighbors from lipid components and water within a distance of 4 Å around each side chain. Normalized solvation profiles of individual residues are plotted in Figure 5 for Dyn A(1-13) and Dyn A(2-17) from the last 2-2.5 ns of the simulations. As anticipated, the C-terminal residues that are in the interface region interact with water and lipid headgroups in both simulations. In Dyn A(2-17), the N-terminal hydrophobic residues are predominantly solvated by the acyl chains. In Dyn A(1-13), in addition to the acyl chain interactions, these residues (especially Tyr-1) interact with lipid headgroups and water. Thus the presence of tyrosine in the first position of N-terminus brings the helix closer to the membranewater interface in Dyn A(1-13). The solvation pattern of residues in Dyn A(1-13) is very similar to that of Dyn A(1-17) observed in our previous simulations.25 The basic residues, and especially Arg-6 and Arg7, participate in “snorkel-model”40 type interactions in both simulations. The long side chains of arginines make hydrophobic interactions with the acyl chains, and the polar guanidinium group at the end interacts with the headgroup region of lipids and water. This pattern of interactions was observed as well in the simulations of the parent peptide.25 b. Aromatic Residues within the Bilayers. The density profiles showed that Phe-4 is close to the center of the bilayer (Figure 4). The solvation profile of Dyn A(1-13) demonstrated that Tyr-1 is exposed to all components of lipid headgroups and water in addition to acyl chains (Figure 5). Both analyses exhibit the tendency of these two aromatic residues to point in opposite directions within the bilayer. We have plotted the MD trajectories of CR and Cζ atom positions of Tyr and Phe residues along the bilayer normal (Figure 6). For comparison, data from the simulations25 on the native peptide Dyn A(1-17) are also shown. The plots show that the Cζ of Tyr-1 and Phe-4 were close to the lipid headgroups and bilayer center, respectively.

Figure 5. Normalized solvation of peptide side chains plotted for Dyn A(1-13) (in A) and Dyn A(2-17) (in B). The number of nearest neighbors around each side chain within 4 Å was counted, and the contributions from lipid components and water were normalized with respect to the total number of atoms in the side chains. Solvation profiles were computed and averaged for the structures saved over the last 2 ns (Dyn A(1-13)) and 2.5 ns (Dyn A(2-17)) of production runs.

The hydrophobic nature of Phe is much more evident in Dyn A(2-17) in which the ring crosses the center of the bilayer. Interaction Energy Analysis. The density profiles and solvation patterns of basic residues suggested that these have strong interactions with the lipid headgroup region and water (see above). The interaction energies of the basic residues with lipids and water were evaluated as described in our earlier simulation studies25 (Table 1), and their contributions to the peptide-lipid and peptide-water interaction energies were analyzed. For Dyn A(2-17), more than 80% of peptide’s interactions with lipid and 60% of the interactions with water in Dyn A(2-17) come from the basic residues. In Dyn A(113), nearly 90% of peptide’s interactions with lipid and water are contributed by arginines and lysines. Since the hydrophilic C-terminus part is truncated in Dyn A(1-13), the electrostatic interactions are primarily from the basic residues that interact strongly with the phosphate and carbonyl groups of the lipid headgroup region and also with water. The environment of Arg-6 and Arg-7 is shown in Figure 7 for both Dyn A(1-13) and Dyn A(2-17). As seen also in Dyn A(1-17) simulations,25 these interactions play a significant role in keeping the Nterminal helical segments within the bilayers throughout the simulations. Interactions between the basic residues and the phosphate oxygens of the headgroups have also been found in several high-resolution structures of integral membrane proteins such as the photosynthetic reaction center, bacteriorhodopsin,

214 J. Phys. Chem. B, Vol. 106, No. 1, 2002

Sankararamakrishnan and Weinstein

Figure 6. Center of mass location along the bilayer normal for the CR (left) and Cζ (right) atoms of Tyr (gray) and Phe (black) residues. Analysis of (A) Dyn A(1-17),25 (B) Dyn A(1-13), and (C) Dyn A(2-17) show that Tyr-1 and Phe-4 point in opposite directions, toward the lipid headgroup region and bilayer center, respectively. Note the significantly deeper penetration of the bilayer by Phe of Dyn A(2-17) compared to the other two peptides.

cytochrome c oxidase, and the β-barrel proteins from the outer membrane of Gram negative bacteria.41 Water Penetration in the Bilayer. Experimental studies using the phenomenon of wavelength-selective fluorescence have indicated that water penetration does occur in the deep hydrocarbon region of the bilayer in the fluid phase membrane.42 Fluorescence spectroscopy and neutron diffraction studies showed that the presence of membrane-bound peptide or proteins shifts the water distribution toward the hydrophobic interior of the bilayer.43,44 Simulation studies also demonstrated

that the presence of peptide within the bilayers facilitates the water penetration.25,45,46 The solvation profiles of Dyn A(113) also showed that Tyr-1 interacts with water (Figure 5). In contrast, the N-terminal residues of Dyn A(2-17) are predominantly exposed to the lipid acyl chains. The influence of tyrosine residue in attracting the water molecules into the bilayer is quantified by calculating the average number of water molecules within 5 Å of each residue for Dyn A(1-13) and Dyn A(2-17) (Figure 8a; data for Dyn A(1-17)25 are also plotted for comparison). In Dyn A(1-17)

Dynorphin Peptides in Membrane Bilayers

J. Phys. Chem. B, Vol. 106, No. 1, 2002 215

TABLE 1: Average Interaction Energies (kcal/mol) of Peptide and the Basic Residues with Lipids and Water interaction energy

Dyn A(2-17)b

Dyn A(1-13)c

peptide-DMPC Arg/Lys-DMPCa peptide-water Arg/Lys-watera

-789.3 -671.6 (85%) -544.8 -368.2 (68%)

-677.6 -613.5 (90%) -523.5 -463.1 (88%)

a The percentage contributions of basic residues toward the total peptide interaction energies are given in the brackets. b Calculated for the last 2.5 ns of the production run. c Calculated for the last 2 ns of the production run.

and Dyn A(1-13), Tyr-1 interacts with 4-6 water molecules, while the N-terminal hydrophobic residues hardly attract any water molecules. The residues starting from Leu-5 attract fewer water molecules in Dyn A(2-17) compared to Dyn A(1-17)/ Dyn A(1-13). The overall penetration of water molecules in the Dyn A(2-17) system is about 2-3 times smaller than that observed for either Dyn A(1-17) or Dyn A(1-13) (Figure 8b). This can be attributed due to the deeper penetration of the N-terminal helix in Dyn A(2-17) that is due to the absence of Tyr-1. Water penetration is similar for bilayers including Dyn A(113) and Dyn A(1-17). As observed also in Dyn A(1-17) simulations,25 fewer water molecules penetrate the hydrophobic face of the helix that forms about one-third of the total helix surface. Dynorphin A(1-17) in Water. Snapshots of the peptide structures from the simulation of the native peptide Dyn A(117) in water (see Methods) (Figure 9) show that the helical conformation collapses within the first 1 ns simulation after the restraints were removed; the peptide assumes a somewhat

random conformation. This agrees with the FTIR and 1H NMR deuterium exchange studies that suggested an interwoven “unordered” structure.20 Our simulation results show that this random conformation in water is due to various interactions between the hydrophilic residues and water. This is in contrast to the MD simulations of the same peptide or its fragments in DMPC bilayers where the helix conformation between residues 4 and 9 remained stable in four different simulations [two earlier simulations on Dyn A(1-17)25 and the two present simulations on Dyn A(1-13) and Dyn A(2-17)]. These results support the view that the interactions in the phospholipid bilayer membrane are key for stabilizing a defined conformation of imbedded peptide hormones.16-19 Discussion The simulations have revealed the discrete interactions that provide a structural basis for the observed similarity between the physicochemical and hence biological properties of Dyn A(1-17) and Dyn A(1-13), as well as for the surprisingly distinct differences from Dyn A(2-17). Moreover, the comparison of structure and dynamics of Dyn A(1-17) in water to the results from the bilayer simulations suggests that these differences are strongly dependent on the effect of the phospholipid bilayer on the properties of the peptides. In particular, the interactions of Tyr-1 with water and the phospholipid headgroups emerge as key factors in the stabilization and positioning of the peptide in the membrane bilayer. Notably, it is the set of intermolecular interactions, rather than any intramolecular stabilization that explains the determinant role of Tyr-1. Consequently, both the physicochemical and the

Figure 7. “Snorkel-model” type interactions observed for Arg-6 (left) and Arg-7 (right) in Dyn A(1-13) (top) and Dyn A(2-17) (bottom). The structures are from the end of the production runs and identify all interactions within 5 Å of the Arg side chains.

216 J. Phys. Chem. B, Vol. 106, No. 1, 2002

Figure 8. (A) Average number of water molecules around each side chain in the N-terminal segment of dynorphin peptides. (B) Total number of water molecules around the N-terminal helical segments (residues 1-10) plotted for the last 1 ns of the production runs. Data for Dyn A(1-13) are in red and for Dyn A(2-17) in blue. For comparison purposes, analysis from the two simulations25 of Dyn A(117) are also shown (black).

Figure 9. Snapshots for the MD simulation of Dyn A(1-17) in water. At the end of 1.4 ns simulation, the positional restraints on CR atoms were removed. During the subsequent 1 ns simulation, the structure became random.

biological properties of these peptides were found to be determined specifically by the membrane environment in which they act. Notably, the simulations have revealed mechanistic details that are not evident from structure alone. Thus the model of membrane-bound dynorphin derived by Schwyzer suggested that

Sankararamakrishnan and Weinstein the aromatic rings of tyrosine and phenylalanine will be on the same side of the helix.14 This gave rise to the possibility that the helix may be stabilized by stacking interactions between these two aromatic residues. However, we find that in the membrane environment Tyr-1 and Phe-4 are on the opposite sides of the helix due to the different affinities of their side chains for different components of the lipids. While Tyr-1 preferred to be close to the lipid headgroups, Phe-4 is stabilized near the center of the bilayer. The distances between the centers of the aromatic rings in Dyn A(1-17)/Dyn A(1-13) simulations are in the range 9-11 Å (significantly larger than in Schwyzer’s dynorphin model14). The findings from our detailed analysis show that the two aromatic residues behave differently within the bilayers because the differences in the chemical nature of side chains determine preferences for different regions of the bilayer. Interestingly, this observation is in agreement with results from the recent experimental studies47,48 using “minimal glycosylation distance” (MGD). From these studies, Tyr and Trp along with the basic residues were shown to have a strong preference for well-defined positions near the lipid headgroup region. The hydrophobic Phe, in turn, was shown to prefer the transmembrane region. That our results are generalizable is indicated by the preferential positioning of Tyr and Trp residues in the interface region identified in crystal structures of multispanning integral membrane proteins,48,49 apparent in transmembrane domains that are in either R-helical or β-sheet conformations.50 This positioning is also evident from the recent analysis of 29 structures of integral membrane proteins with resolution 4 Å or better.51 The residue distribution of Tyr obtained from these structures shows peaks near the interface, while Phe is distributed throughout the transbilayer region. Statistical preference of Trp and Tyr for the interface regions is observed as well in the analysis of single-spanning proteins.52,53 Our studies clearly show that the basic residues participate in a special type of interactions in the interface region, termed “snorkel model”. These special interactions are most likely to help the peptide remain within the bilayers and determine its orientation.25 Such special interactions of positively charged residues with the membrane interface have been suggested recently from experimental studies.48 The presence of lipids at the protein-membrane interface has been identified in several high-resolution integral membrane protein structures, and these lipids have been shown to interact with the basic residues.41 Moreover, analysis of three-dimensional crystal structures of membrane proteins indicated that the basic residues are excluded from the central region of transmembrane helical segments and showed that they occur predominantly in the interface.51 In attempts to connect structural properties to mechanisms, a variety of functional roles have been proposed for aromatic residues present at the interface between the lipid and the polar region of the membrane.54-56 These include suggestions that these residues play a stabilizing role as interfacial anchoring residues or that they are determinants of protein orientation.48,57 In the present simulation studies on Dyn A(1-17)/Dyn A(113), the primary role of tyrosine seems to be to keep the peptide close to the interface through its specialized interactions with the lipid headgroup and water. This positional role played by Tyr-1 may have significant mechanistic implications for the biological properties of peptide. In the case of dynorphin, the absence of Tyr-1 residue that results in deeper penetration of Dyn A(2-17) may explain the significantly different biological activity because the peptide lacking the Tyr may be prevented from recognizing the opioid receptor that is the normal target

Dynorphin Peptides in Membrane Bilayers of the (1-17) peptide. Such insights are of practical significance because the opioid actions of dynorphin made it a potential candidate for a clinical analgesic, with fewer adverse side effects than are associated with the classic opiate morphine.58 Recently, a controlled pilot study has been carried out on morphine-treated chronic pain patients to test the analgesic efficacy and tolerability of Dyn A(1-13).59 This study concluded that a brief intravenous infusion of Dyn A(1-13) could potentially augment analgesia in opioid-treated patients. Although substitutions and modifications of many dynorphin residues revealed their importance in the activity and potency,5,60 the most dramatic effect occurred when the first residue Tyr was removed.61 The orientation of the peptide inside the bilayers is also determined by characteristic interactions of the Tyr-1 and Phe-4 residues in Dyn A(1-17)/Dyn A(1-13). Our detailed analysis showed how these interactions are responsible for positioning the peptides in a tilted orientation with respect to the bilayer normal. With Tyr-1 removed, the force pulling the peptide upward from its N-terminus toward the lipid headgroup is eliminated, resulting in the deeper penetration of the peptide Dyn A(2-17) into the bilayer. This is likely to be a generalized property of peptide-membrane interactions and is testable experimentally. Thus it was shown recently that such conclusions about the depth of peptide penetration can be probed with electron spin resonance62,63 and low-angle diffraction techniques.64,65 Using the property of paramagnetic influence of oxygen on a fluorine nucleus, 19F NMR was used to determine the membrane immersion depth of an antibacterial peptide indolicidin.66 Given the differences identified here from the MD simulations, it is likely that for opioid peptides as for many other biologically active peptides, the interaction mechanisms that determine the position and orientation within the membrane will be important determinants of their biological properties. Because these mechanisms depend on identifiable physicochemical properties, as shown here, the type of analysis illustrated in the present study is likely to yield important insight into functional affinity and specificity of the peptides for their specific targets, such as the G-protein coupled receptors residing in the membrane of the cell. Given this perceived importance, it must be noted that information about the effects of embedded peptides and proteins on the physical and chemical properties of membranes is still scant. The present simulations of Dyn A(1-17) and its fragments utilized carefully considered experimental parameters such as lipid cross-sectional area and bilayer thickness as well as rather lengthy simulations by current standards (up to 9 ns). Clearly, it is difficult to obtain a complete sampling of peptide/lipid interactions from a single MD simulation, but our work has already demonstrated the use of several different starting points for simulations of the same system.25 Notably, the same final results were obtained in that case. It is noteworthy, as well, that due to the use of an NVE ensemble simulation in which the box lengths are fixed, there remains the unanswered question whether the entire effect of the peptide on the physical properties of the membrane has been accounted for. Our current simulations use an NPγΤ ensemble representation with Dyn A(1-13) and a larger cutoff of 18 Å to address this question. Initial observation of the 4 ns production trajectory reveals that the peptide adopts an orientation similar to the one observed in the present simulations. Acknowledgment. This work was supported by NIH grants P01 DA-11470, DA-12923, and K05 DA-00060. Computational support was provided by the Cornell Supercomputer Facility and the Advanced Scientific Computing Laboratory at the

J. Phys. Chem. B, Vol. 106, No. 1, 2002 217 Frederick Cancer Research Facility of the National Cancer Institute (Laboratory for Mathematical Biology). References and Notes (1) Goldstein, A.; Fischli, W.; Lowney, L. I.; Hunkapillar, M.; Hood, L. Proc. Natl. Acad. Sci. U.S.A. 1981, 78, 7219. (2) Chavkin, C.; Goldstein, A. Proc. Natl. Acad. Sci. U.S.A. 1981, 78, 6543. (3) Chavkin, C.; James, I. F.; Goldstein, A. Science 1982, 215, 413. (4) Corbett, A. D.; Paterson, S. J.; McKnight, A. T.; Magnan, J.; Kosterlitz, H. W. Nature 1982, 299, 79. (5) Naqvi, T.; Haq, W.; Mathur, K. B. Peptides 1998, 19, 1277. (6) Tang, Q.; Lynch, R. M.; Porreca, F.; Lai, J. J. Neurophysiol. 2000, 83, 2610. (7) Hooke, L. P.; He, L.; Lee, N. M. J. Pharmacol. Exp. Ther. 1995, 273, 802. (8) Caudle, R. M.; Mannes, A. J. Pain 2000, 87, 235. (9) Tang, Q.; Gandhoke, R.; Burritt, A.; Hruby, V. J.; Porreca, F.; Lai, J. J. Pharmacol. Exp. Ther. 1999, 291, 760. (10) Hiramatsu, M.; Inoue, K. Brain Res. 2000, 859, 303. (11) Itoh, H.; Andoh, T.; Watanabe, I.; Sasaki, T.; Kamiya, Y.; Okumura, F. Eur. J. Neurosci. 2000, 12, 1253. (12) Hooke, L. P.; He, L.; Lee, N. M. J. Pharmacol. Exp. Ther. 1995, 273, 292. (13) Schwyzer, R. Biopolymers 1991, 31, 785. (14) Schwyzer, R. Biopolymers (Peptide Sci.) 1995, 37, 5. (15) Moroder, L.; Romano, R.; Guba, W.; Mierke, D. F.; Kessler, H.; Delporte, C.; Winand, J.; Christoph, J. Biochemistry 1993, 32, 13551. (16) Inooka, H.; Ohtaki, T.; Kitahara, O.; Ikegami, T.; Endo, S.; Kitada, C.; Ogi, K.; Onda, H.; Fujino, M.; Shirakawa, M. Nature: Struct. Biol. 2001, 8, 161. (17) Whitehead, T. L.; McNair, S. D.; Hadden, C. E.; Young, J. K.; Hicks, R. P. J. Med. Chem. 1998, 41, 1497. (18) Bader, R.; Bettio, A.; Beck-Sickinger, A. G.; Zerbe, O. J. Mol. Biol. 2001, 305, 307. (19) Rovero, P.; Pellegrini, M.; Di Fenza, A.; Meini, S.; Quartara, L.; Maggi, C. A.; Formaggio, F.; Toniolo, C.; Mierke, D. F. J. Med. Chem. 2001, 44, 274. (20) Renugopalakrishnan, V.; Rapaka, R. S.; Huang, S.-G.; Moore, S.; Hutson, T. B. Biochem. Biophys. Res. Commun. 1988, 151, 1220. (21) Lancaster, C. R. D.; Mishra, P. K.; Hughes, D. W.; St.-Pierre, S. A.; Bothner-By, A. A.; Epand, R. M. Biochemistry 1991, 30, 4715. (22) Spadaccini, R.; Crescenzi, O.; Picone, D.; Tancredi, T.; Temussi, P. A. J. Peptide Sci. 1999, 5, 306. (23) Tessmer, M. R.; Kallick, D. A. Biochemistry 1997, 36, 1971. (24) Erne, D.; Sargent, D. F.; Schwyzer, R. Biochemistry 1985, 24, 4261. (25) Sankararamakrishnan, R.; Weinstein, H. Biophys. J. 2000, 79, 2331. (26) Brooks, B. R.; Bruccoleri, R. E.; Olafson, B. D.; States, D. J.; Swaminathan, S.; Karplus, M. J. Comput. Chem. 1983, 4, 187. (27) Mackerell, A. D., Jr.; Bashford, D.; Bellot, M.; Dunbrack, R. L.; Evanseck, J. D.; Field, M. J.; Fischer, S.; Gao, J.; Guo, H.; Ha, S.; JosephMcCarthy, D.; Kuchnir, L.; Kuczera, K.; Lau, T. K.; Mattos, C.; Michnick, S.; Ngo, T.; Nguyen, D. T.; Prodhom, B.; Reiher, W. E.; Roux, B.; Schlenkrich, B.; Smith, J.; Stote, R.; Straub, J.; Watanabe, M.; WiorkiewiczKuczera, J.; Karplus, M. J. Phys. Chem. B 1998, 102, 3586. (28) Schlenkrich, M.; Brickmann, J.; Mackerell, A. D., Jr.; Karplus, M. Biological Membranes: A molecular perspectiVe from computation and experiment; Merz, K. M., Jr., Roux, B., Eds.; Birkhauser: Boston, 1996; Chapter 2. (29) Jorgensen, W. L.; Chandrasekhar, J.; Madura, J. D.; Impey, R. W.; Klein, M. L. J. Chem. Phys. 1983, 79, 926. (30) Gennis, R. Biomembrane: Molecular Structure and Function; Springer-Verlag: New York, 1989. (31) Jakobsson, E.; Subramaniam, S.; Scott, H. L. Biological Membranes: A molecular perspectiVe from computation and experiment; Merz, K. M., Jr., Roux, B., Eds.; Birkhauser: Boston, 1996; Chapter 4. (32) Chiu, S.-W.; Subramaniam, S.; Jakobsson, E. Biophys. J. 1999, 76, 1929. (33) Woolf, T. B.; Roux, B. Proteins: Struct. Func. Genet. 1996, 24, 92. (34) Berneche, S.; Nina, M.; Roux, B. Biophys. J. 1998, 75, 1603. (35) Roux, B.; Woolf, T. B. Biological Membranes: A molecular perspectiVe from computation and experiment; Merz, K. M., Jr., Roux, B., Eds.; Birkhauser: Boston, 1996; Chapter 17. (36) Nagle, J. F.; Tristram-Nagle, S. Biochim. Biophys. Acta 2000, 1469, 159. (37) Venable, R. M.; Zhang, Y.; Hardy, B. J.; Pastor, R. W. Science 1993, 262, 223. (38) Woolf, T. B.; Roux, B. J. Am. Chem. Soc. 1994, 116, 5916. (39) Sankararamakrishnan, R.; Konvicka, K.; Mehler, E. L.; Weinstein, H. Int. J. Quantum Chem. 2000, 77, 174.

218 J. Phys. Chem. B, Vol. 106, No. 1, 2002 (40) Segrest, J. P.; DeLoof, H.; Dohlman, J. G.; Brouillette, C. G.; Anantharamaiah, G. M. Proteins: Struct. Func. Genet. 1990, 8, 103. (41) Fyfe, P. K.; McAuley, K. E.; Roszak, A. W.; Isaacs, N. W.; Cogdell, R. J.; Jones, M. R. Trends Biochem. Sci. 2001, 26, 106. (42) Chattopadhyay, A.; Mukherjee, S. J. Phys. Chem. B 1999, 103, 8180. (43) Ho, C.; Stubbs, C. D. Biophys. J. 1992, 63, 897. (44) Jacobs, R. E.; White, S. H. Biochemistry 1989, 28, 3421. (45) Ha Duong, T.; Mehler, E. L.; Weinstein, H. J. Comput. Phys. 1999, 151, 358. (46) Bachar, M.; Becker, O. J. Chem. Phys. 1999, 111, 8672. (47) Braun, P.; von Heijne, G. Biochemistry 1999, 38, 9778. (48) Killian, J. A.; von Heijne, G. TIBS 2000, 25, 429. (49) von Heijne, G. Prog. Biophys. Mol. Biol. 1996, 66, 113. (50) White, S. H. Membrane proteins of known three-dimensional structure, http://blanco.biomol.uci.edu/Membrane_Proteins_xtl.html. (51) Ulmschneider, M. B.; Sansom, M. S. P. Biochim. Biophys. Acta 2001, 1512, 1. (52) Landolt-Marticorena, C.; Williams, K. A.; Deber, C. M.; Reithmeier, R. A. F. J. Mol. Biol. 1993, 229, 602. (53) Arkin, I. T.; Brunger, A. T. Biochim. Biophys. Acta 1998, 1429, 113.

Sankararamakrishnan and Weinstein (54) Moosmann, B.; Behl, C. Eur. J. Biochem. 2000, 267, 5687. (55) White, S. H.; Wimley, W. C. Annu. ReV. Biophys. Biomol. Struct. 1999, 28, 319. (56) White, S. H. Biophys. J. 2001, 80, Part 2, 5a. (57) Schiffer, M.; Chang, C. H.; Stevens, F. J. Protein Eng. 1992, 5, 213. (58) Millan, M. J. Trends Pharmacol. Sci. 1990, 11, 70. (59) Portenoy, R. K.; Caraceni, A.; Cherny, N. I.; Goldblum, R.; Ingham, J.; Inturrisi, C. E.; Johnson, J. H.; Lapin, J.; Tiseo, P. J.; Kreek, M. J. Clin. Drug InVest. 1999, 17, 33. (60) Hruby, V. J.; Agnes, R. S. Biopolymers (Peptide Sci.) 1999, 51, 391. (61) Walker, J. M.; Moises, H. C.; Coy, D. H.; Baldrighi, G.; Akil, H. Science 1982, 218, 1136. (62) Altenbach, C.; Greenhalgh, D. A.; Khorana, H. G.; Hubbell, W. L. Proc. Natl. Acad. Sci. U.S.A 1994, 91, 1667. (63) Hubbell, W. L.; Mchaourab, H. S.; Altenbach, C.; Lietzow, M. A. Structure 1996, 4, 779. (64) McIntosh, T. J. Curr. Top. Membr. 1999, 48, 23. (65) Lohner, K.; Prenner, E. J. Biochim. Biophys. Acta 1999, 1462, 141. (66) Prosser, R. S.; Luchette, P. A.; Westerman, P. W.; Rozek, A.; Hancock, R. E. W. Biophys. J. 2001, 80, 1406.