Solvation of Transmembrane Proteins by Isotropic Membrane

J. Phys. Chem. B 2007, 111, 11285-11296

11285

Solvation of Transmembrane Proteins by Isotropic Membrane Mimetics: A Molecular Dynamics Study Madhusoodanan Mottamal, Sui Shen, Cristina Guembe, and Goran Krilov* Department of Chemistry, Boston College, 2609 Beacon Street, Chestnut Hill, Massachusetts 02467 ReceiVed: May 24, 2007; In Final Form: June 27, 2007

Mixtures of organic solvents are often used as membrane mimetics in structure determination of transmembrane proteins by solution NMR; however, the mechanism through which these isotropic solvents mimic the anisotropic environment of cell membranes is not known. Here, we use molecular dynamics simulations to study the solvation thermodynamics of the c-subunit of Escherichia coli F1F0 ATP synthase in membrane mimetic mixtures of methanol, chloroform, and water with varying fractions of components as well as in lipid bilayers. We show that the protein induces a local phase separation of the solvent components into hydrophobic and hydrophilic layers, which provides the anisotropic solvation environment to stabilize the amphiphilic peptide. The extent of this effect varies with solvent composition and is most pronounced in the ternary methanol-chloroform-water mixtures. Analysis of the solvent structure, including the local mole fraction, density profiles, and pair distribution functions, reveals considerable variation among solvent mixtures in the solvation environment surrounding the hydrophobic transmembrane region of the protein. Hydrogen bond analysis indicates that this is primarily driven by the hydrogen-bonding propensity of the essential Asp61 residue. The impact of the latter on the conformational stability of the solvated protein is discussed. Comparison with the simulations in explicit all-atom models of lipid bilayer indicates a higher flexibility and reduced structural integrity of the membrane mimetic solvated c-subunit. This was particularly true for the deprotonated form of the protein and found to be linked to solvent stabilization of the charged Asp61.

I. Introduction The thermodynamic stability of protein conformations and hence their biological activity critically depend on the interactions of the polypeptide chain with its solvent medium. Although the effects of aqueous solvent on structure of water-soluble proteins have been extensively studied,1-6 due to relative dearth of structural data, our understanding of the solvation in cell membranes is much less complete. Transmembrane proteins function in a complex physiological environment consisting of a hydrophobic domain (the lipid bilayer) bounded by aqueous domains on either side. The partitioning of hydrophobic and hydrophilic residues between these environments dominates the thermodynamics of folding of transmembrane proteins and to a large extent determines their global structure. As a consequence of this pronounced amphiphilicity, it is often very difficult to obtain crystals of transmembrane proteins of sufficient quality to allow high-resolution X-ray structural determination. Hence, a large fraction of structural studies relies on solution NMR. However, NMR measurements in lipid bilayer or membrane-mimicking detergent micelles are hindered by line broadening due to long correlation times and overlap of the signals from the protein and membrane-forming medium. The latter often leads to a significant degradation in the quality of NOE signals, making unambiguous assignment of molecular structure difficult. To avoid these difficulties, high-resolution NMR studies of transmembrane proteins are frequently performed in organic mixtures that function as isotropic membrane mimetics. In many cases, there is evidence that transmembrane proteins dissolved in the above solvents fold into conformations closely * To whom correspondence should be addressed. E-mail: [email protected].

resembling their native membrane bound structures.7-10 Escherichia coli protein EmrE was found to retain its biological activity in methanol-chloroform mixtures.11 R-Helical segments of bacteriorhodopsin were found to have similar structures when solvated in both biphasic SDS micelles and monophasic methanol-chloroform mixtures.7 More recently, the structure of the E. coli c-subunit of F1F0 ATP synthase was determined in a methanol-chloroform-water mixture and found to be in good agreement with in situ mutant studies.12,13 Hence, structural data obtained in isotropic membrane mimetics are often used to provide insight into biochemical processes that occur in membranes in ViVo. Nonetheless, the mechanism through which such isotropic mixtures mimic the highly anisotropic solvation environment found in cell membranes is not well understood. The latter is significant, because the biological function of transmembrane proteins is often strongly influenced by the interactions of the peptide with the lipid bilayer. For example, binding to membrane surfaces was found to catalyze the recognition between peptide hormones and their receptors.14,15 Hence, to be able to draw meaningful conclusions about their in ViVo biochemistry from solution NMR data, it is essential to improve the understanding of membrane mimetic solvents and their effect on the structure and dynamics of amphiphilic proteins. Molecular dynamics simulations have been indispensable in the study of microscopic structure and dynamics of liquids.16-20 In contrast with the large body of work on aqueous solutions, there have been comparatively few studies of solvation in simple organic mixtures.21-24 Mixtures of methanol and chloroform are among the most commonly used isotropic membrane mimetics due to their exceptional efficiency in solubilizing amphiphilic

10.1021/jp0740245 CCC: $37.00 © 2007 American Chemical Society Published on Web 09/05/2007

11286 J. Phys. Chem. B, Vol. 111, No. 38, 2007

Mottamal et al.

TABLE 1: OPLS 2001 Potential Parameters for Methanol and Chloroform Lennard Jones Parameters atom

charge

σ [Å]

O C HO HC

-0.683 0.145 0.418 0.040

Methanol 3.12 3.50 0.00 2.50

0.170 0.066 0.000 0.030

C Cl H

0.083 -0.062 0.103

Chloroform 3.50 3.47 2.50

0.066 0.266 0.030

[kcal/mol]

Bond Stretching Parameters atom

Kr [kcal/mol/Å2]

req (Å)

O-C O-HO C-HC

Methanol 1.410 0.945 1.090

320.00 553.37 340.23

C-Cl C-H

Chloroform 1.781 1.090

245.16 340.23

Angle Bending Parameters atom type

Kθ [kcal/mol/deg2]

θeq [deg]

O-C-HC HO-O-C HC-C-HC

Methanol 109.5 108.5 107.8

35.02 55.04 33.02

Cl-C-Cl H-C-Cl

Chloroform 107.6 111.7

78.05 51.03

Torsional Parameters atom type

V1 [kcal/mol]

Hc-C-O-Ho

1.10

V2 [kcal/mol]

V3 [kcal/mol]

Methanol 0.0

0.45

molecules.7,9,12,13,25,26 Although their components are miscible in all proportions, both experimental and computational studies have shown evidence of self-clustering in methanol-chloroform mixtures, leading to microscopic heterogeneity.27-29 Simulations of small amphiphilic molecules dissolved in these mixtures indicate that this tendency leads to preferential solvation of the hydrophilic and hydrophobic ends by the methanol and chloroform respectively.29 Preferential solvation has also been shown to enhance the propagation of π-bulges in the simulations of the unfolding of a transmembrane segment of bacteriorhodopsin.30 In this paper, we use molecular dynamics simulations to investigate the microscopic mechanism through which isotropic membrane mimetics stabilize transmembrane proteins. In particular, we chose to study the solvation dynamics of c-subunit of the E. coli F1F0 ATP synthase, an enzyme that utilizes the flow of protons through the cell membrane to power a rotary catalytic machinery that enables the synthesis and hydrolysis of cellular ATP.13,31-33 The c-subunit received considerable attention recently and is believed to play a crucial role in coupling the transmembrane proton flux to the catalytic rotation of the enzyme assembly.13,31-36 Recent NMR measurements of the E. coli c-subunit in a 4:4:1 methanol-chloroform-water mixture by Girvin et al.12 have shown that the peptide folds in hairpin motif with two membrane spanning helical segments connected by a polar loop. Moreover, two distinct stable conformations of the protein were observed at low pH and high

Figure 1. Snapshot of the final configuration of the protein-solvent system following a 5 ns simulation with the protein backbone restrained showing the solvent components methanol (gold), chloroform (bluegreen), and water (yellow) around the c-subunit (bright green ribbons). A and B show the solvent distribution around the protonated and deprotonated form of the c-subunit in the ternary 4:4:1 methanolchloroform-water mixture, whereas C and D show the solvent distribution around the protonated and deprotonated form of the protein in the binary 1:1 methanol-chloroform mixture.

pH. These results form the basis of a widely accepted model of the rotary mechanism, which postulates that the rotation of the enzyme assembly is driven by conformational changes of the c-subunit induced by local variations in pH.13,31 On the other hand, newest NMR measurements on a closely related thermophilic bacillus PS3 c-subunit in a 1:3 methanol-chloroform mixture show that the protein folds into a single structure that is independent of pH variations.37 Hence, a detailed understanding of the solvent contribution to the stabilization of the c-subunit conformations and its variation as a function of pH and solvent composition is of particular interest. Elucidating the extent to which membrane mimetics are able to reproduce the solvation environment of cell membranes is equally important in qualifying solution NMR data. We investigate this by comparing the structural properties of the c-subunit extracted from the simulations in mixtures of methanol, chloroform, and water with the results of the simulations of the protein embedded in a lipid bilayer. Although the crystal structure of the c-subunit of E. coli F1F0 ATP synthase is not known, the high-resolution structure of the c-subunit of related enzymes, the I. tartaricus and E. hirae Na+ATP synthase, has been solved by X-ray diffraction.38,39 Although the charge carriers of the two enzymes differ, the overall fold of the c-subunits from the two enzymes is expected to be similar. This paper is organized as follows. In section II, we describe the simulation methods and discuss the solvation of the c-subunit in several isotropic membrane mimetics. In section III, we address the conformational stability of the c-subunit in lipid bilayers, and in section IV, we discuss important proteinsolvent interactions. We conclude in section V. II. Subunit c in Membrane Mimetic Solvents A. Simulation Details. We used all-atom molecular dynamics simulations to study the solvation of the c-subunit of F1F0 ATP

Solvation of Transmembrane Proteins

J. Phys. Chem. B, Vol. 111, No. 38, 2007 11287

TABLE 2: Composition of Membrane Mimetic Solvents Used in Molecular Dynamics Simulations of Solvation of Protonated and Deprotonated c-Subunit Systems mixture

c-subunit

methanol

chloroform

water

Na+

4:4:1 methanol-chloroform-water

protonated deprotonated protonated deprotonated protonated deprotonated protonated deprotonated

2850a 2859 1429 1427 4266 4259 2844 2844

2843 2838 4270 4273 1418 1415 2839 2843

716 715 0 0 0 0 0 0

1 2 1 2 1 2 1 2

1:3 methanol-chloroform 3:1 methanol-chloroform 1:1 methanol-chloroform a

The values indicate the number of molecules of each species.

synthase by four different isotropic membrane mimetic mixtures: (i) 4:4:1 methanol-chloroform-water mixture (similar to the solvent used by Girvin et al.12 in the NMR study of the E. coli c-subunit), (ii) 1:1 methanol-chloroform mixture, which lacks water but has equal ratios of organic solvents as in the previous case, (iii) 1:3 methanol-chloroform mixture, used by Akutsu and co-workers in the NMR study of bacillus PS3 c-subunit,37 and (iv) 3:1 methanol-chloroform mixture, where the ratio of organic solvents were reversed. All simulations were performed using the GROMACS package.40,41 The potential parameters for methanol and chloroform were taken from the OPLS 2001 force field,42 whereas the SPC model43 was used for water. The assigned potential parameters for methanol and chloroform are shown in Table 1. In all membrane mimetic systems, initial configuration of the solvent mixtures was chosen to be a cubic lattice of L ) 90 Å generated from a unit cell containing the appropriate solvent components in the aforementioned ratios. This assured optimal a priori mixing of solvents in all four solvent systems. Initial structures were relaxed via a 2000 step steepest descent (SD) energy minimization. The systems were then equilibrated via a 500 ps molecular dynamics (MD) simulation using the Leapfrog integrator with a time step of 1 fs. Berendsen44 thermostat and barostat were used to maintain constant temperature (298 K) and pressure (1 atm), and periodic boundary conditions were used throughout. Electrostatic interactions were treated using the particle mesh Ewald (PME) method45,46 with a grid spacing of 0.08 nm and a cutoff distance of 1.4 nm. In all equilibration simulations, the density of the system was found to converge after ∼200 ps. The starting c-subunit configurations for our simulations were obtained from the low pH (PDB code: 1c0v) and high pH (PDB code: 1c99) solution NMR structures published by Girvin and co-workers.12,13 Essential Asp61 carboxyllate was assumed to be fully protonated at pH 5 and fully deprotonated at pH 8.31 All other titrable residues were treated as either protonated or deprotonated depending on their pKa values. The simulation systems were prepared as follows. The protein was inserted in the center of the pre-equilibrated box of membrane mimetic solvent mixture. Solvent molecules overlapping with the protein were deleted, and the system was neutralized by adding an appropriate number of counterions. Table 2 shows the final number of solvent molecules used in the simulations of the protonated and deprotonated forms of the protein in all four membrane mimetic solvents. The protein-solvent systems were equilibrated via a 1000 step SD energy minimization to remove bad contacts, followed by 500 ps MD simulation at constant NPT holding the protein backbone restrained. The harmonic restraining potential with force constant k ) 1000 kJ/(mol nm2) was applied to all CR atoms. Room temperature and pressure were maintained using Berendsen thermostat and barostat,44 and periodic boundary conditions were applied in all directions.

Long-range electrostatics was treated by the particle mesh Ewald method45,46 with a grid spacing of 0.08 nm and a cutoff distance of 1.4 nm. The equations of motion were integrated using the Leapfrog algorithm47 with a time step of 1 fs. Two canonical production simulations were then performed for each system. First, the protein backbone was restrained, and a 5 ns trajectory was generated to explore the structure and dynamics of the solvent in the vicinity of the protein. The restraints were then removed, and the trajectory was evolved for a further 5 ns to study the stability of the protein structure in these membrane mimetic mixtures. In both simulations, Nose-Hoover thermostats48 were used to maintain temperature, and full periodic boundary conditions were used throughout. Configurations were saved in 1 ps intervals, and data from the final 4 ns of each simulation was used in analysis. B. Solvent Structure and Solvation Mechanism. We first analyzed the solvent behavior around the NMR structure of the c-subunit in the four membrane mimetic mixtures, using the data from the restrained stage of the simulation. Final configurations, depicting the solvent distribution around the protonated and deprotonated c-subunit in a 4:4:1 methanol-chloroformwater system following 5 ns of restrained simulation are shown in Figure 1A and B, respectively. The transmembrane peptide can be divided into three distinct regions, based on its native solvation environment: the hydrophilic helix termini, the hairpin loop, and hydrophobic transmembrane helix segments. The solvent distribution around the solute protein displayed in Figure 1A and B reveals a marked difference from that expected for a homogeneous mixture. In both cases, the membrane mimetic mixture is found to partition in the vicinity of the protein in such a way that the polar termini and loop regions of the protein are well solvated almost exclusively by the polar component (water and methanol), whereas the hydrophobic transmembrane region of the peptide is solvated primarily by the low dielectric nonpolar solvent, chloroform. Hence, the amphiphilic protein induces a local phase separation of the solvent mixture, with the resulting trilayer structure resembling the lipid-water interface of cell membranes. Thus, the water/methanol rich phase surrounding the termini and loop takes on the role of the layer populated by water and charged lipid headgroups, and the chloroform rich region takes the function of the hydrophobic tail region of the lipid bilayer. On the other hand, inspection of the distribution of the solvent components in the simulation of protein solvation in the closely related anhydrous solvent, the 1:1 methanol-chloroform mixture (Figure 1C and D) indicates that in the absence of water, the extent of the local phase separation was reduced significantly, particularly for the deprotonated c-subunit. This is further illustrated in Figure 2, which shows the layer mole fraction profile of solvent components, defined as:


∫ Fl(x,y,z)dxdy

Xl(z) )

∑k ∫ Fk(x,y,z)dxdy

Mottamal et al.

(1)

where z is the principal axis of the protein roughly parallel to the two helices, and Fk(x,y,z) is the number density of the k-th component of the solution. In the 4:4:1 methanol-chloroformwater solution (Figure 2A), the mole fraction of chloroform is significantly elevated in the vicinity of the hydrophobic transmembrane region of the protein, whereas the methanol and water fractions dominate the layers adjacent to the polar termini and loop regions. The crossover points from methanol/water-rich to chloroform-rich layers agree well with the expected boundaries between hydrophobic and hydrophilic segments of the protein sequence. A similar behavior is observed in layer mole fractions of solvent components in anhydrous membrane mimetic mixtures as well, shown in Figure 2B-D. Although transitions between hydrophobic and hydrophilic layers are clearly distinguishable for all three mixtures, the phase separation is incomplete. In particular, the partitioning of the solvent components in 3:1 and 1:1 methanol-chloroform mixtures (Figure 2C and D) is considerably less pronounced than in case of the 1:3 methanol-chloroform mixture (Figure 2B), indicating that membrane-mimicking properties are composition-dependent. Figure 3 shows the local mole fraction of the solvent components as a function of distance from the principal axis (aligned with the z-axis) for each of the three regions of the peptide with distinct solvation characteristics. The local mole fraction was computed as:

∫zz

i+1

Xl(r) )

dz

i

∑k ∫z

zi+1

i

∫02π Fl(r,φ,z)rdφ

dz

∫0

2π

(2) Fk(r,φ,z)rdφ

where Fk(r,φ,z) is the number density of k-th component of the solvent. The protein was divided along the z-axis into three regions zi < z < zi+1: the hydrophilic loop [z0 < z < z1], the hydrophobic TMH region [z1< z < z2], and the polar termini [z2 < z < z3]. The region boundaries zi were chosen based on biochemical information.49 This characterization is due to approximate cylindrical symmetry of the protein and clearly defined solvent layer boundaries observed in Figure 2. We note that the mole fractions computed by eq 2 do not include contributions from the solvent molecules extending beyond the extrema of the loop and termini (i.e., those with z < z0 and z > z3). However, as shown in Figure 2, the distribution of the solvent molecules in this “cap” regions is identical to that in the respective loop and termini regions described by eq 2. Hence, the omission of these end effects does not appreciably alter the solvation profiles of the polar regions of the protein. In the case of 4:4:1 methanol-chloroform-water mixture (Figure 3A), both protonated and deprotonated peptides show an increased concentration of methanol and water in the polar loop and termini regions. Conversely, for r > 10 Å, the concentration of chloroform is elevated in the hydrophobic TMH regions. This corresponds well to the expected peptide-chloroform contact distance, which can be estimated as the sum of the helix diameter and chloroform molecular radius to be ∼11 Å. This is true for anhydrous solvents as well, but the effect is considerably weaker in 3:1 and 1:1 methanol-chloroform mixtures (Figure 3C and D). Unlike the ternary 4:4:1 mixture where the departure of solvent fractions from random mixing

Figure 2. Layer mole fraction of the solvent components methanol (black line), chloroform (red line), and water (green line) measured along the principal axis of the protein. The results are shown for the protonated (top panels) and deprotonated (bottom panels) forms of the c-subunit in a 4:4:1 methanol-chloroform-water mixture (A), and 1:3 (B), 3:1 (C), and 1:1 (D) methanol-chloroform mixtures.

was observed even at considerable distances (greater than 20 Å) from the protein, separation of the solvent components in 3:1 and 1:1 methanol-chloroform mixtures was more localized. The solvent partitioning in these systems was clearly defined only in the immediate vicinity of the protein and rapidly approached the distribution of the pure solvent mixture as the distance from the protein increased. On the other hand, bulk distribution was recovered for the 4:4:1 mixture only at substantially larger distances, ∼45 Å (data not shown). To further characterize the structure of the solvent in the vicinity of the protein, in Figure 4 we plot the radial density profiles of the solvent components defined as:

Fl(r) )

1 F0,l

∫zz

i+1

i

dz

∫02π Fl(r,φ,z)rdφ

(3)

where zi denote the boundaries between the three regions specified above. The profiles for methanol and water in the vicinity of the polar loop and termini regions are more structured than those for chloroform, especially for the protonated c-subunit in the 4:4:1 methanol-chloroform-water mixture (Figure 4A). This was also true for methanol in anhydrous solvent mixtures, where most pronounced solvent ordering was observed in the 1:3 mixture (Figure 4B), followed by the 1:1 mixture (Figure 4D). In the 3:1 mixture (Figure 4C), methanol was less structured in both polar regions of the peptide. However, all three anhydrous solvents were found to be less structured than the ternary 4:4:1 mixture, especially away from the peptide. The ordering of the solvent components is primarily due to hydrogen bonding between hydrophilic residues and the polar solvent components, whereas chloroform, which mostly solvates the lipid-exposed region of the protein through nonspecific hydrophobic interactions, is less structured. One feature of interest is a sharp peak in the water density profile at r ) 2 Å for the TMH region of deprotonated c-subunit in the ternary mixture. We believe this is due to particularly strong hydrogen bonding between the charged Asp61 and small number of surface water molecules in an otherwise mostly hydrophobic environment. This is illustrated in Figure 5A, which shows a snapshot of the typical solvation environment in the vicinity of the abovementioned charged residue. On the other hand, the sharp water



Figure 3. Local mole fraction of solvent components methanol (black line), chloroform (red line), and water (green line) as a function of distance from the principal axis of the protein. The mole fractions were computed separately for the solvation environment around each of the three distinct regions of the c-subunit: the polar loop (top panels), hydrophobic TMH (middle panels), and polar termini (bottom panels). The results are shown for the protonated (left panels) and deprotonated (right panels) forms of the c-subunit in a 4:4:1 methanol-chloroform-water mixture (A), and 1:3 (B), 3:1 (C), and 1:1 (D) methanol-chloroform mixtures.

peak is not present in the density profiles of the solvent components for the TMH region of the corresponding protonated system. This is consistent with the radial mole fraction data in Figure 4A, which indicates that the short-range water population is essentially nonexistent in this region. Instead, one finds that the water signal has been replaced by a weaker methanol peak at r ) 2.5 Å. Figure 5B indicates that the solvation environment in the vicinity of Asp61 of the protonated c-subunit consists almost exclusively of methanol molecules. Hence, the protonated Asp61 does not provide sufficient hydrogen bonding affinity to induce substantial migration of water into the hydrophobic layer. It is plausible that this qualitative difference in the local solvation environment of the protonated and deprotonated form of the c-subunit results in different protein conformations being stabilized under acidic or basic conditions, as observed by Girvin et al.12 This interpretation is also consistent with the results of Nakano et al.,37 who found no structural variation between protonated and deprotonated forms of the thermophilic bacillus PS3 c-subunit in anhydrous 1:3 methanol-chloroform mixtures, as well as the findings of Dimroth and co-workers,50 who observed no significant differences in NMR spectra of the Na+

transducting P. modestum c-subunit at pH 5.8, 7.0 and 7.5 in 4:4:1 methanol-chloroform-water mixtures. In summary, partial local phase separation of the solvent components in the vicinity of the protein was observed in all four membrane mimetic mixtures, but the effect in anhydrous solvents was more localized. Comparison of c-subunit solvation in 1:3, 1:1, and 3:1 methanol-chloroform mixture shows that the extent of local phase separation is composition dependent and decreases as methanol content is increased. To our knowledge, this is the first atomic level demonstration of the mechanism through which isotropic membrane mimetics stabilize transmembrane proteins. III. Solvation of Subunit c in Lipid Bilayers A. Molecular Dynamics Simulations. To evaluate the quality of the structure and the stability of the folded protein in membrane mimetic solvents, both the protonated and deprotonated form of the E. coli c-subunit were simulated in the lipid bilayer environment. All the simulations were performed using the NAMD package51 with protein and lipid interaction param-


Mottamal et al.

Figure 4. Cylindrical density profiles of solvent components methanol (black line), chloroform (red line), and water (green line) as a function of distance from the principal axis of the protein. The density profiles were computed separately for the solvation environment around each of the three distinct regions of the c-subunit: the polar loop (top panels), hydrophobic TMH (middle panels), and polar termini (bottom panels). The results are shown for the protonated (left panels) and deprotonated (right panels) forms of the c-subunit in a 4:4:1 methanol-chloroform-water mixture (A), and 1:3 (B), 3:1 (C), and 1:1 (D) methanol-chloroform mixtures.

Figure 5. Representative local solvation environment around charged (A) and protonated (B) Asp61 residues of the c-subunit from the in the 4:4:1 methanol-chloroform-water mixture. Water molecules are observed as the primary hydrogen-bonding partners of the charged Asp61 (A), whereas methanol molecules are the primary hydrogen-bonding partners of the protonated residue (B).

eters taken from CHARMM-27 force field.52 The bilayer was constructed of commonly used palmitoyloleoyl-phosphatidylcholine (POPC) phospholipids using the membrane builder in the VMD program,53 and both sides of the bilayer were soaked with ∼15 Å deep layers of TIP3P water.54 The protein was inserted with its principal axis orthogonal to the membrane, such that the hydrophobic helices were embedded in the lipid bilayer, whereas the polar termini and loop regions were immersed in water. All lipid molecules overlapping with the protein were carefully removed. Both systems were neutralized by adding the corresponding number of counterions. The final compositions of the two simulation systems are summarized in Table 3. Initial dimension of the simulation box for both systems was approximately 67 × 65 × 82 Å. The membrane-protein systems were relaxed via a two-stage energy minimization procedure using a combination of the conjugate gradient and line search algorithms, consisting of a 2000 step minimization with all protein backbone atoms fixed, followed by 3000 step minimization with CR atoms restrained by a harmonic restraining potential with a force constant k ) 1.0 kcal/(mol Å2). Equilibra-



TABLE 3: Parameters of the Systems Used in Simulations of Solvation of the c-subunit in Lipid Bilayer c-subunit

POPC

water

Na+

box size [Å]

protonated deprotonated

91a 89

4412 4492

1 2

56 × 54 × 83 56 × 54 × 82

a

The values indicate the number of molecules of each species.

tion was achieved in three stages. Starting with the minimized structure, the CR atoms were restrained, and the system was evolved for 150 ps using Langevin dynamics with a damping coefficient of 10 ps-1. This was followed by a 600 ps MD simulation at constant pressure and temperature. The restraints were then removed, and the system was evolved for a further 600 ps. Pressure was maintained at 1 bar using the Nose-Hover Langevin dynamics with a piston period of 200 fs and a piston decay of 100 fs, whereas Langevin thermostats were used to maintain the temperature at 300 K. Equations of motion were integrated with a time step of 1 fs. Electrostatic interactions were computed by the particle mesh Ewald method, with grid density 1 Å-1, and periodic boundary conditions were applied in all directions. Maintaining the lipid packing consistent with that of the native membrane is essential for accurate representation of the physical properties of the lipid bilayer in molecular dynamics simulations. Following equilibration, the average area per lipid was calculated for both protonated and deprotonated systems using the Voronoi tessellation method55 and found to be 62 Å2, which is very close to the accepted experimental value of 63 Å2.56 Production runs were performed for 10 ns at constant temperature and pressure, with reduced Langevin damping constant (1 ps-1) and slower piston decay time (500 fs). In addition, the average area per lipid was held constant to help maintain the integrity of the membrane. This was achieved by allowing only those cell fluctuations which keep the xy area constant. The coordinates were saved every 2 ps, and structural and dynamic information was extracted from the last 8 ns of the simulations. B. Structural Stability and Conformational Dynamics of the Solvated Protein. The structures of the protonated and deprotonated c-subunit in the lipid bilayer following the 10 ns unrestrained simulations are shown in Figure 6A and B, respectively. The transmembrane region of the protein is primarily solvated by the hydrophobic tail of the lipid, whereas the polar termini and loop regions are well solvated by the polar head groups of the lipids and the surrounding water molecules. Backbone RMSDs of the final structure from the original NMR structures were found to be relatively small at 2.4 and 3.0 Å for protonated and deprotonated c-subunit, respectively. In contrast, the RMS deviations in best performing membrane mimetic solvents (4:4:1 methanol-chloroform-water and 1:3 methanol-chloroform mixtures) was observed to be close to 4.5 Å following 5 ns of simulation, whereas backbone RMSD of the peptide in methanol-rich membrane mimetic mixtures (1:1 and 3:1 methanol-chloroform) was around 8 Å. Thus, the peptide in the lipid bilayer was found to be significantly more stable than in membrane mimetic systems. Furthermore, the stability of the peptide in membrane mimetic solvents showed strong dependence on the ratio of polar to nonpolar components in the mixture. Our study indicates that sufficient amount of nonpolar solvent is essential for adequate solvation of the long transmembrane helices in membrane mimetic solvents. Clearly defined hydrophobic/hydrophilic partitioning of the lipid-water interface and greater rigidity due to slow diffusion of the lipids may contribute to the greater structural stability of the protein in the bilayer system. In contrast, in the membrane mimetic system, any conformational change of the peptide can easily

Figure 6. Snapshot of the final configuration of the protonated (A) and deprotonated (B) form of the c-subunit following a 10 ns unrestrained simulation in POPC lipid bilayer.

be accommodated through rapid rearrangement of the mobile solvent molecules to stabilize the new structure. Because the solvent molecules surrounding the polar termini and loop and the nonpolar hydrophobic TMH regions of the protein in the membrane mimetic solutions are much smaller than the head group or the hydrophobic tail of the lipid molecules, the organic solvents can pack well around every region of the protein including the interhelical region, thereby making the peptide more flexible than in the lipid bilayer. Figure 7A and B show the backbone RMSD of protonated c-subunit from the NMR structure as a function of time in the explicit lipid bilayer and membrane mimetic environments, respectively. RMSDs are calculated for the whole protein, the loop region, and the transmembrane helix region separately. In the lipid bilayer, backbone RMSD of the whole peptide was observed to steadily increase for the first 2 ns and then plateau, fluctuating between 2 and 3 Å for the remainder of the 10 ns simulation. This indicates that, following minor adjustments from the NMR structure, the protein conformation is overall stable. In contrast, the fluctuation in the backbone RMSD of the peptide in the membrane mimetic solvent was significantly larger than in the lipid bilayer. The RMSD values in these systems were found to oscillate between 2.5 and 5 Å throughout the simulation, indicating greater flexibility, and therefore lower structural stability than observed in a lipid bilayer environment. RMSD analysis of different regions of the peptide shown in Figure 7A and B shows that in the lipid bilayer the loop region exhibits larger structural fluctuations than either of the transmembrane helices, whereas in the membrane mimetic solvent, the polar loop region shows less fluctuation than the transmembrane helices. The former is understandable, because in the functioning enzyme the c-subunit loops are most likely further stabilized through interaction with the γ unit.57,58 Hence, we can conclude that the structural integrity of the hairpin fold is better preserved in the lipid bilayer, primarily due to stabilization of the TMH region, whereas in the membrane mimetic solvent it is the loop region that is better stabilized by the solvent, leading to the overall reduced stability of the hairpin fold. This is also evident from the plots of per residue RMSD shown in Figure 7C and D for the protonated c-subunit in lipid bilayer and membrane mimetic environments, respectively. In the lipid bilayer, largest fluctuations are observed for residues 38-44, which constitute the loop region, whereas in the membrane mimetic solvents, most significant values of the RMSD are observed for residues that populate the two helices. In addition, termini regions of the peptide in both systems also showed large fluctuations due to their innate flexibility.


Mottamal et al.

Figure 7. Backbone RMSD of the protonated form of the c-subunit. (A) and (B) RMSD as a function of time for the whole protein (black), TMH-1 (red), TMH-2 (green), and the loop region (blue) in the POPC lipid bilayer and the 4:4:1 methanol-chloroform-water mixture, respectively. (C) and (D) Average RMSD per residue for these two solvation environments. Residues 4-38 and 49-76 correspond to transmembrane helices, whereas residues 39-48 comprise the loop.

Paralleling our analysis of the solvation in membrane mimetic mixtures, lipid solvation around three regions of the peptide was examined by considering the average position of residues in the lipid bilayer. For this purpose, we divided the lipid layers into the head group region and the hydrophobic tail region. Head groups of the lipids are defined as the centers of mass of the phosphate and choline groups in each leaflet. Boundaries of the head group region in each leaflet are delimited by: ΣZHgcom/n ( σ ; where ZHgcom are the z coordinates of the centers of mass of the head groups and σ is the standard deviation. The region in between the head groups in the upper and lower leaflet is treated as the hydrophobic part of the bilayer. Positions of amino acid residues are specified by the centers of mass of backbone atoms. Figure 8 shows the percent hydrophobicity profile of the protonated and deprotonated c-subunit in the lipid bilayer obtained from the 10 ns production run. Percent hydrophobicity of the residue is determined by the fraction of time a residue spends in the head group or in hydrophobic region of the bilayer. Amino acid residues 11-31 and 53-74 in both protonated and deprotonated systems were found to reside almost exclusively in the hydrophobic region of the lipid bilayer, whereas the residues comprising the polar loop and termini are almost always found in the head group or in the aqueous region. This result indicates that the protein is firmly anchored in the membrane and is in agreement with the experimental data which suggest

Figure 8. Percent hydrophobicity of the protonated (red) and deprotonated (black) c-subunit measured as a function of the residence time of the center of mass of each residue in the hydrophobic lipid tail region of the bilayer.

that amino acid residues 10-31 and 54-76 are the membrane spanning segments of the transmembrane helices.49 The two-solution NMR structures by Girvin et al.12 for the protonated and deprotonated form of the c-subunit differ



Figure 9. Configurational fluctuations of the protonated form (top panels) and deprotonated form (bottom panels) of the c-subunit in the POPC lipid bilayer. (A) Bending angle of TMH-1 (black) and TMH2. (B) Crossing angle of the upper (red) and lower (black) segments of the transmembrane helices. (C) Relative twist angle of the upper (red) and lower (black) segments of the transmembrane helices.

significantly. Whereas the two helices are parallel in the protonated structure, in the deprotonated structure the two transmembrane helices have kinks around residues 26 and 60 in TMH-1 and TMH-2, respectively.12 More significantly, TMH-2 is rotated around its principal axis by approximately 140° with respect to TMH-1.12 To characterize the stability of these structures in the lipid bilayer at different protonation state, we monitored the conformational fluctuations including the helical bending, crossing, and twist angles. Helical bending was characterized in the following fashion. Each transmembrane helix was partitioned into an upper and lower helix with the boundary between the two at residues 26 and 60 for TMH-1 and TMH-2, respectively. Helical axis was defined by the centers of mass of the CR atoms of 4 consecutive amino acid residues from each end of the helix (upper and lower helices were defined separately). The angle formed by the upper and lower helical axes was defined as the bending angle. Similarly, the angle between the two upper helices or the two lower helices was measured as the crossing angle. The twist angle specifies the degree of rotation of TMH-1 with respect to TMH-2 around the principal helix axis. This order parameter is of interest as the major structural difference between the protonated and deprotonated forms of the c-subunit observed by Girvin and co-workers12,13 involves changes in the twist angle. The orientation of each helix was specified by a vector perpendicular to the helical axis and passing through a reference CR atom. Figure 9A, B, and C show the bending, crossing, and twist angles, respectively, for the protonated and deprotonated proteins as a function of time. For the protonated protein, average bending angles were found to be 10 and 15° for TMH-1 and TMH-2, respectively, whereas the same angles in the deprotonated protein were found to be significantly larger: 28 and 52° for TMH-1 and TMH-2, respectively. These values are close to those observed in the NMR structures.12 The average crossing angle between the upper helices that connect the loop is ∼7°, whereas the same between the lower helices, which are free at the N and C termini is 20° in the protonated c-subunit. Thus, the region of the TMHs attached to the loop is more rigid than the region of transmembrane helices that are part of the N and C termini. Average crossing angles for the upper and lower helices are almost identical (∼16°) in the deprotonated system. This suggests that both the upper and lower segments of the TMHs are equally flexible in the deprotonated c-subunit. This flexibility is possible due to the presence of a proline kink in TMH-2 and also due to the presence of flexible loop and termini

Figure 10. Hydrogen bonding with Asp61. (A) and (B) Number of hydrogen bonds as a function of time with methanol (black) and water (red) for the protonated and deprotonated form of the c-subunit in the 4:4:1 methanol-chloroform-water mixture. (C) Existence of the hydrogen bond with the CR-H of Ala24 for the deprotonated c-subunit in the POPC lipid bilayer.

at both ends of the membrane spanning helices. Average twist angle of upper helix of TMH-1 with respect to the upper helix of TMH-2 in the protonated system is ∼8°, whereas the same for the lower helix is ∼12°. Similarly in the deprotonated system, average twist angles are 4° and 9° for the upper and lower helices, respectively. In the protonated system, fluctuation of twist angles of the lower helices is only slightly larger than that of the upper helices. In contrast, fluctuations of twist angles in the lower helices of the deprotonated protein are much larger than that of the upper helices. This again shows the increased flexibility of lower TMHs that are parts of the free N and C terminal ends, which is further enhanced by helix bending. In general, structural fluctuations are more pronounced in the deprotonated system than in the protonated system. The structural features of the c-subunit in a lipid bilayer are generally in agreement with the NMR data of Girvin et al.12 In particular, the increased flexibility observed in the deprotonated form is likely to facilitate helical bending leading to kinks found in the high pH NMR structure.12 IV. Characterization of Important Protein-Solvent Interactions The penetration of methanol and water into the chloroform rich regions such as that observed in snapshots of simulations in membrane mimetic solvents (Figure 1) suggests that the hydrophobic transmembrane regions of deprotonated c-subunit have greater affinity for polar components than the transmembrane regions of protonated peptide. Moreover, the polar solvent population in the hydrophobic layer appears to be strongly localized in the proximity of Asp61, especially in the deprotonated system. This implies that stronger affinity of the charged Asp61 for polar components enhances deeper penetration of the latter into the hydrophobic layer than the protonated Asp61. Further evidence of this can be found by analyzing the hydrogen-bonding pattern of the solvent components to Asp61. Figure 10 shows the number of hydrogen bonds between Asp61 and polar solvents in the 4:4:1 methanol-chloroform-water membrane mimetic solvents, as well as the number of hydrogen


Figure 11. Orientation of Asp61 and Ala24 in the protonated (A) and deprotonated (B) c-subunit in POPC lipid bilayer.

bonds between Asp61 and the proximal Ala24 residue in the lipid bilayer, as a function of time. For the protonated c-subunit, Asp61 forms on average 1.65 hydrogen bonds with the polar solvent components, with methanol molecules as the primary hydrogenbonding partner comprising 80% of the total hydrogen bonds, the remaining bonds being with water. In contrast, the charged Asp61 of the deprotonated c-subunit is a more efficient hydrogen bonder, forming an average of 4.17 hydrogen bonds with polar components. Moreover, a significantly higher fraction of hydrogen bonds is to water (45%), which is consistent with the large population of water in the first solvation shell of the transmembrane region observed in the local mole fraction data in Figure 3A, as well as sharp water peaks in the cylindrical density profile for water in Figure 4A. On the other hand, simulations of the c-subunit in the lipid bilayer indicate no hydrogen bonding between Asp61 and water. This is expected, as the highly hydrophobic environment of the lipid bilayer is very unfavorable for water. Similarly, although we do not find any interhelix hydrogen bonds in the protonated protein, in the deprotonated form of the c-subunit, the charged Asp61 was found to form a special CR-H‚‚‚O hydrogen bond with Ala24 of TMH-1. Numerous transmembrane proteins containing GXXXG and GXXXA motifs are known to be stabilized by CR-H‚‚‚O bonding. Quantum mechanical calculations predict the energy of a CR-H‚‚‚O bond in model systems to be 2.5-3.0 kcal/mol,59,60 and FTIR measurements estimate the bond energy of CR(G79)-H‚‚‚O(I76) in glycophorin A to be 0.88 kcal/mol.61 However, this energy can vary considerably depending upon the arrangement of the interacting groups.62 Sequence analysis of the c-subunit shows a GXXXG (G23AAIG27) motif in TMH-1 and a GXXXA (G58LVDA62) motif in TMH-2. Because these two motifs are in close proximity, we analyzed the environment around Asp61 in the lipid bilayer system for both protonated and deprotonated cases. Figure 11A and B show a typical orientation of the key residues Ala24 and Asp61 from the two motifs in the protonated and deprotonated systems, respectively. In the protonated system, side chain of Asp61 was oriented away from the helix-helix interface, whereas in the deprotonated system, the side chain of Asp61 was oriented toward the interface of the two helices. In addition, the carboxyl oxygen of Asp61 was oriented toward the CR-H of Ala24, in a configuration favorable to hydrogen bonding. This is consistent with our analysis of interhelix hydrogen bonds in Figure 10C, which shows that the CR-H‚‚‚O hydrogen bond is present in 39% of the sampled configurations, whereas no interhelix hydrogen bonds were detected for the protonated protein. In both cases, geometric criteria of CR-O distance < 3.8 Å63 and angle ∠O‚C-H < 30° were used to indicate the presence of a hydrogen bond. In contrast, the solution NMR structure of the deprotonated c-subunit, as well as our membrane mimetic solvent simulations

Mottamal et al. show the distance between Ala24-CR and Asp61-OD to be consistently larger than 5.2 Å, precluding the possibility of interhelix hydrogen bonding. Rather, the stabilization of the charged Asp61 was through hydrogen bonding with polar solvent components penetrating the hydrophobic chloroform layer. Because Asp61 is embedded in a low dielectric environment, stronger affinity of charged Asp61 to polar solvents can be attributed to the weak dielectric screening of the charge. In summary, both in membrane mimetic mixtures and the lipid bilayer, hydrogen bonding interactions with Asp61 play an important role in facilitating the conformational stability of the c-subunit monomer, however the nature of these interactions in the two environments is quite distinct. It is therefore very plausible that the variations in the local solvation environment around the essential Asp61 residue may strongly affect the relative stability of the c-subunit conformers. Hence, interpretation of the mechanisms of in ViVo processes based solely on the changes observed in membrane mimetic solution structures may be difficult. V. Conclusions Although organic solvent mixtures are widely used as membrane mimetics, relatively little is known about the mechanism through which they mimic the highly anisotropic solvation environment of the lipid-water interface to stabilize amphiphilic transmembrane proteins. In the present study, we have addressed this issue through molecular dynamics simulations of solvation of a representative transmembrane protein, the c-subunit of E. coli F1F0 ATP synthase, in four membrane mimetic solvent mixtures. Examination of the solvent structure in the vicinity of the c-subunit showed strong tendency for preferential solvation of the hydrophilic loop and termini regions, and the hydrophobic transmembrane helix regions by the polar and nonpolar solvent components, respectively. This leads to substantial local demixing of the solvents and partial phase separation. This effect was most pronounced for the threecomponent methanol-chloroform-water mixture. In binary methanol-chloroform mixtures the local phase separation is less pronounced, with the effect being the strongest for 1:3 methanolchloroform mixtures, and weakening with the increasing fraction of methanol. Furthermore, in all binary mixtures, the phase separation was localized to the immediate vicinity of the protein, where the effect extended and persisted over shorter distances from the protein, than in case of ternary mixtures. It is therefore not coincidental that two of the solvent mixtures showing the highest tendency for local phase separation, the 4:4:1 methanolchloroform-water and 1:3 methanol-chloroform, are those that have been successfully used in solution NMR structural studies of the c-subunit from E. coli12 and bacillus PS3.37 The variations observed among different membrane mimetic mixtures indicates that the demixing propensity is governed by a delicate balance of hydrophilic and hydrophobic interactions between the solvent components on one side, and the solute protein on the other. The prominent role of hydrogen bonding is evident by comparing the solvation properties of the 4:4:1 methanol-chloroform-water mixture with the closely related 1:1 methanol-chloroform mixture. Although both contain equal proportions of methanol and chloroform, through its ability to form extensive hydrogen bond networks, water is able to promote extensive aggregation of the polar components in the 4:4:1 mixture. In contrast, self-aggregation in methanol is driven by entanglement of short hydrogen-bonded chains, which is less effective, and hence the entropic cost of demixing limits the cluster size.

Solvation of Transmembrane Proteins Solvation properties were also found to depend on the protonation state of the essential titrable Asp61 residue, and hence the pH of the solution. In the deprotonated system, the charged Asp61 promotes substantial penetration of polar solvent components into the chloroform rich region surrounding the hydrophobic helices. This allows the charged residue to be stabilized through hydrogen bonding in the first solvation shell. Although structural differences between different solvent mixtures can be observed in all three regions around the c-subunit protein, they are most pronounced in hydrophobic helix region of the deprotonated peptide. In the ternary mixture, water is found to displace methanol as the primary hydrogen-bonding partner of Asp61. This certainly affects the solvation free energy of Asp61 and may well have implications on the relative stability of the protonated and deprotonated conformations. In particular, it may explain the differences between the structure of the deprotonated c-subunit observed by solution NMR in the 4:4:1 methanol-chloroformwater mixture by Girvin et al.12 and that found by Nakano et al.37 in the 1:3 methanol-chloroform mixture, as well as by Matthey et al. in 4:4:1 methanol-chloroform-water mixture.50 Comparison of the RMSDs of the protein from the solution NMR structure in membrane mimetic solvents and lipid bilayer showed that the protein is significantly more flexible in the membrane mimetic solution. This is primarily due to differences in the fluidity and relative hydrophobicity/hydrophilicity of the two environments. In the organic solvents, structural fluctuations were mainly in the transmembrane helices, whereas in the lipid bilayer the loop region was found to be more flexible. In general, structural fluctuations are more prominent in the deprotonated protein than in the protonated protein, and largely centered around Asp61. In contrast with the membrane mimetic mixtures, where the charged Asp61 was well solvated by the polar solvent components, in the lipid bilayer Asp61 was primarily stabilized via CR-H‚‚‚O hydrogen bonding to Ala24. This type of CRH‚‚‚O hydrogen bonds are commonly found in membrane proteins with GXXXG and/or GXXXA motifs and are known to stabilize the protein by ∼1 kcal/mol per hydrogen bond. Interestingly, the conformational transition upon the deprotonation of Asp61 proposed by Girvin et al.12 based on their solution NMR data would be necessary to position the above two residues in an orientation favoring CR-H‚‚‚O formation. Still, since the structure of the c-subunit of E. coli ATP synthase in lipid bilayer is not available, the importance of the observed CR-H‚‚‚O hydrogen in stabilizing the protein within the c-ring oligomer of the functioning enzyme cannot be fully evaluated. Nonetheless, this study shows that hydrogen bonding to Asp61 is significant for the conformational stability of the c-subunit monomer in both membrane mimetic and lipid bilayer solvation environments. In summary, we found that the solvation properties of membrane mimetic solvents are strongly composition dependent. In particular, the addition of water to the organic solvent mixtures enhances the local phase separation of solvents, and thus provides a better membrane mimetic compared to binary mixtures of methanol and chloroform. The ratio of hydrophilic and hydrophobic mixture components is also important, with chloroform rich mixtures yielding better solvation properties. Significant differences in the solvation profiles of various membrane mimetic solvents and the lipid bilayer indicate that direct comparisons and conclusions drawn based on solution structures of transmembrane proteins with processes in ViVo may be complicated.

J. Phys. Chem. B, Vol. 111, No. 38, 2007 11295 Acknowledgment. This work was supported by a grant from Boston College to G.K. We would like to thank M. Roberts for helpful discussions. References and Notes (1) Rhee, Y. M.; Sorin, E. J.; Jayachandran, G.; Lindahl, E.; Pande, V. S. Proc. Natl. Acad. Sci. 2004, 101, 6456. (2) Brooks, C. L.; Karplus, M. J. Mol. Biol. 1989, 208, 159. (3) Onuchic, J. N.; LutheySchulten, Z.; Wolynes, P. G. Annu. ReV. Phys. Chem. 1997, 48, 545. (4) Cheung, M. S.; Garcia, A. E.; Onuchic, J. N. Proc. Natl. Acad. Sci. 2002, 99, 685. (5) Papoian, G. A.; Ulander, J.; Eastwood, M. P.; Luthey-Schulten, Z.; Wolynes, P. G. Proc. Natl. Acad. Sci. 2004, 101, 3352. (6) Meyer, E. Protein Sci. 1992, 1, 1543. (7) Pervushin, K. V.; Orekhov, V. Y.; Popov, A. I.; Musina, L. Y.; Arseniev, A. S. Eur. J. Biochem. 1994, 219, 571. (8) Orekhov, V. Y.; Pervushin, K. V.; Korzhnev, D. M.; Arseniev, A. S. J. Biomol. NMR 1995, 6, 113. (9) Dmitriev, O.; Jones, P. C.; Jiang, W. P.; Fillingame, R. H. J. Biol. Chem. 1999, 274, 15598. (10) Engler, A.; Stangler, T.; Willbold, D. Eur. J. Biochem. 2002, 269, 3264. (11) Schwaiger, M.; Lebendiker, M.; Yerushalmi, H.; Coles, M.; Groger, A.; Schwarz, C.; Schuldiner, S.; Kessler, H. Eur. J. Biochem. 1998, 254, 610. (12) Girvin, M. E.; Rastogi, V. K.; Abildgaard, F.; Markley, J. L.; Fillingame, R. H. Biochemistry 1998, 37, 8817. (13) Rastogi, V. K.; Girvin, M. E. Nature 1999, 402, 263. (14) Romano, R.; Dufresne, M.; Prost, M. C.; Bali, J. P.; Bayerl, T. M.; Moroder, L. Biochim. Biophys. Acta 1993, 1145, 235. (15) Moroder, L.; Romano, R.; Guba, W.; Mierke, D. F.; Kessler, H.; Delporte, C.; Winand, J.; Christophe, J. Biochemistry 1993, 32, 13551. (16) Breed, J.; Sankararamakrishnan, R.; Kerr, I. D.; Sansom, M. S. P. Biophys. J. 1996, 70, 1643. (17) Geissler, P. L.; Dellago, C.; Chandler, D.; Hutter, J.; Parrinello, M. Science 2001, 291, 2121. (18) Jorgensen, W. L.; Madura, J. D. J. Am. Chem. Soc. 1983, 105, 1407. (19) Venables, D. S.; Schmuttenmaer, C. A. J. Chem. Phys. 2000, 113, 3249. (20) Saiz, L.; Padro, J. A.; Guardia, E. J. Phys. Chem. B 1997, 101, 78. (21) Kovacs, H.; Kowalewski, J.; Laaksonen, A. J. Phys. Chem. 1990, 94, 7378. (22) Schoen, M.; Hoheisel, C. Mol. Phys. 1986, 58, 699. (23) Michael, D.; Benjamin, I. J. Chem. Phys. 2001, 114, 2817. (24) Bernardi, E.; Stassen, H. J. Chem. Phys. 2004, 120, 4860. (25) Szyperski, T.; Vandenbussche, G.; Curstedt, T.; Ruysschaert, J. M.; Wuthrich, K.; Johansson, J. Protein Sci. 1998, 7, 2533. (26) Maslennikov, I. V.; Arseniev, A. S.; Tchikin, L. D.; Kozhich, A. T.; Bystrov, V. F.; Ivanov, V. T. Biol. Mem. 1991, 8, 156. (27) Singh, P. P.; Sharma, B. R.; Sidhu, K. S. Can. J. Chem. 1979, 57, 387. (28) Moelwynhughes, E. A.; Missen, R. W. J. Phys. Chem. 1957, 61, 518. (29) Gratias, R.; Kessler, H. J. Phys. Chem. B 1998, 102, 2027. (30) Korzhnev, D. M.; Orekhov, V. Y.; Arseniev, A. S.; Gratias, R.; Kessler, H. J. Phys. Chem. B 1999, 103, 7036. (31) Fillingame, R. H.; Dmitriev, O. Y. Biochim. Biophys. Acta 2002, 1565, 232. (32) Jones, P. C.; Hermolin, J.; Jiang, W. P.; Fillingame, R. H. J. Biol. Chem. 2000, 275, 31340. (33) Stock, D.; Leslie, A. G. W.; Walker, J. E. Science 1999, 286, 1700. (34) Jones, P. C.; Hermolin, J.; Fillingame, R. H. J. Biol. Chem. 2000, 275, 11355. (35) Tsunoda, S. P.; Aggeler, R.; Yoshida, M.; Capaldi, R. A. Proc. Natl. Acad. Sci. 2001, 98, 898. (36) Sambongi, Y.; Iko, Y.; Tanabe, M.; Omote, H.; Iwamoto-Kihara, A.; Ueda, I.; Yanagida, T.; Wada, Y.; Futai, M. Science 1999, 286, 1722. (37) Nakano, T.; Ikegami, T.; Suzuki, T.; Yoshida, M.; Akutsu, H. J. Mol. Biol. 2006, 358, 132. (38) Meier, T.; Polzer, P.; Diederichs, K.; Welte, W.; Dimroth, P. Science 2005, 308, 659. (39) Murata, T.; Yamato, I.; Kakinuma, Y.; Leslie, A. G. W.; Walker, J. E. Science 2005, 308, 654. (40) Berendsen, H. J. C.; Vanderspoel, D.; Vandrunen, R. Comput. Phys. Comm. 1995, 91, 43. (41) Lindahl, E.; Hess, B.; Van Der Spoel, D. J. Mol. Modeling 2001, 7, 306.

11296 J. Phys. Chem. B, Vol. 111, No. 38, 2007 (42) Kaminski, G. A.; Friesner, R. A.; Tirado-Rives, J.; Jorgensen, W. L. J. Phys. Chem. B 2001, 105, 6474. (43) Hermans, J.; Berendsen, H. J. C.; Vangunsteren, W. F.; Postma, J. P. M. Biopolymers 1984, 23, 1513. (44) Berendsen, H. J. C.; Postma, J. P. M.; Vangunsteren, W. F.; Dinola, A.; Haak, J. R. J. Chem. Phys. 1984, 81, 3684. (45) Essmann, U.; Perera, L.; Berkowitz, M. L.; Darden, T.; Lee, H.; Pedersen, L. G. J. Chem. Phys. 1995, 103, 8577. (46) Darden, T.; York, D.; Pedersen, L. J. Chem. Phys. 1993, 98, 10089. (47) Hockney, R. W.; Goel, S. P.; Eastwood, J. W. J. Comput. Phys. 1974, 14, 148. (48) Evans, D. J.; Holian, B. L. J. Chem. Phys. 1985, 83, 4069. (49) Fillingame, R. H.; Jiang, W.; Dmitriev, O. Y.; Jones, P. C. Biochim. Biophys. Acta 2000, 1458, 387. (50) Matthey, U.; Braun, D.; Dimroth, P. Eur. J. Biochem. 2002, 269, 1942. (51) Kale, L.; Skeel, R.; Bhandarkar, M.; Brunner, R.; Gursoy, A.; Krawetz, N.; Phillips, J.; Shinozaki, A.; Varadarajan, K.; Schulten, K. J. Comput. Phys. 1999, 151, 283. (52) MacKerell, A. D.; Bashford, D.; Bellott, M.; Dunbrack, R. L.; Evanseck, J. D.; Field, M. J.; Fischer, S.; Gao, J.; Guo, H.; Ha, S.; Joseph-

Mottamal et al. McCarthy, D.; Kuchnir, L.; Kuczera, K.; Lau, F. T. K.; Mattos, C.; Michnick, S.; Ngo, T.; Nguyen, D. T.; Prodhom, B.; Reiher, W. E.; Roux, B.; Schlenkrich, M.; Smith, J. C.; Stote, R.; Straub, J.; Watanabe, M.; Wiorkiewicz-Kuczera, J.; Yin, D.; Karplus, M. J. Phys. Chem. B 1998, 102, 3586. (53) Humphrey, W.; Dalke, A.; Schulten, K. J. Mol. Graph. 1996, 14, 33. (54) Jorgensen, W. L.; Chandrasekhar, J.; Madura, J. D.; Impey, R. W.; Klein, M. L. J. Chem. Phys. 1983, 79, 926. (55) Shinoda, W.; Okazaki, S. J. Chem. Phys. 1998, 109, 1517. (56) Konig, B.; Dietrich, U.; Klose, G. Langmuir 1997, 13, 525. (57) Andrews, S. H.; Peskova, Y. B.; Polar, M. K.; Herlihy, V. B.; Nakamoto, R. K. Biochemistry 2001, 40, 10664. (58) Schulenberg, B.; Aggeler, R.; Murray, J.; Capaldi, R. A. J. Biol. Chem. 1999, 274, 34233. (59) Vargas, R.; Garza, J.; Dixon, D. A.; Hay, B. P. J. Am. Chem. Soc. 2000, 122, 4750. (60) Scheiner, S.; Kar, T.; Gu, Y. L. J. Biol. Chem. 2001, 276, 9832. (61) Arbely, E.; Arkin, I. T. J. Am. Chem. Soc. 2004, 126, 5362. (62) Mottamal, M.; Lazaridis, T. Biochemistry 2005, 44, 1607. (63) Derewenda, Z. S.; Lee, L.; Derewenda, U. J. Mol. Biol. 1995, 252, 248.

Solvation of Transmembrane Proteins by Isotropic Membrane

Recommend Documents