Molecular dynamics simulations of hydrophobic and amphiphatic

Computational and Theoretical Polymer Science 10 (2000) 97–102

Molecular dynamics simulations of hydrophobic and amphiphatic proteins interacting with a lipid bilayer membrane J.-H. Lin, A. Baumgaertner* Forum Modellierung, Forschungszentrum, 52425 Ju¨lich, Germany Received 19 April 1999; received in revised form 21 June 1999; accepted 27 June 1999

Abstract Molecular dynamics simulations of polypeptides at high dilution near a fully hydrated bilayer membrane have been performed. In contrast to previous theoretical predictions, Monte Carlo simulations and conclusions from experiments a spontaneous insertion of amphiphatic or hydrophobic proteins into a membrane is not observed. Rather it is found that an amphiphatic chain has the tendency to remain in proximity to the membrane surface, whereas the location of a hydrophobic chain is more unbound. This is shown using two proteins, melittin and polyleucine. The conformation of the proteins and their orientation with respect to the membrane surface are discussed. q 2000 Elsevier Science Ltd. All rights reserved. Keywords: Molecular dynamics simulation; Hydrophobicity; Membrane; Proteins

1. Introduction The adhesion at and partitioning of proteins into membranes and their subsequent folding are fundamental processes in biological cells. It has been known for a long time that short proteins, in many cases toxins, adsorb and translocate spontaneously into membranes [1,2]. Despite many experimental facts indicating spontaneous adsorption and insertion under certain experimental conditions, our theoretical understanding of this translocation process is still poor [3–11]. In the present work we report on some results from molecular dynamics simulations of two proteins, melittin and polyleucine, interacting at high dilution with a fully hydrated lipid bilayer. Melittin is 26 amino acids long including the C terminus and the N terminus. It has the sequence (H2N-Gly-Ile-Gly pAla-Val-Leu-Lys p -Val-Leu-Thr-Thr-Gly-Leu-Pro-AlaLeu-Ile-Ser-Trp-Ile-Lys p -Arg p -Lys p -Arg p -Gln-GlnCONH2) where charged amino acids are indicated by an asterisk. Melittin is the major protein component of the venom of the European honey bee Apis mellifera. It has a hemolytic activity. The N terminus part (Ile-Gly-Ala-ValLeu) is more hydrophobic, whereas the anchor sequence

* Corresponding author. Tel.: 1 49-2461-61-4074; fax: 1 49-2461-612983. E-mail address: [email protected] (A. Baumgaertner).

(Lys p-Arg p-Lys p-Arg p-Gln-Gln) is positively charged and hence strongly hydrophilic. Polyleucine (Leu25) consists of 25 identical side chains of the amino acid leucine. The C terminus and the N terminus are acetyl group and amine group, respectively, and therefore this peptide is completely uncharged. Leucine is one of the most hydrophobic amino acids. According to the Roseman scale [12] or the Eisenberg scale [13] the average hydrophobicity of Leu25 is 275 kcal/mol and that of melittin is about 2 13 kcal/mol. This energy scale is the relative free energy change of transport from an aqueous to a nonpolar environment. Since Leu25 and melittin are very hydrophobic, one would expect that these proteins would spontaneously insert into the membrane. This view has been supported by Monte Carlo simulations using a simplified membrane model [4,5,9]. However, as it will be shown by the present work, more detailed atomistic models do not support the evidence of a spontaneous insertion. The main reason of previous shortcomings is probably the inadequate modeling of the water–membrane interface. This interface essentially prevents the proteins to come into close contact with the nonpolar core of the bilayer membrane. It is important to note that our findings are in agreement with observations and experiments on analogs of melittin, like alamethicin [6] and others [14], at very high dilution. At higher concentrations, however, experiments indicate insertion of melittin in a probably aggregated form [15].

1089-3156/00/$ - see front matter q 2000 Elsevier Science Ltd. All rights reserved. PII: S1089-315 6(99)00062-8

98

J.-H. Lin, A. Baumgaertner / Computational and Theoretical Polymer Science 10 (2000) 97–102

Fig. 1. The distance z(t) of the center-of-mass of melittin and polyleucine with respect to the midplane between two adjacent membrane surfaces as a function of time t.

2. Model and simulation techniques The membrane consists of a bilayer of 2 × 100 lipids of the type POPC (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine). Each lipid consists of 134 atoms. Carbon atoms with hydrogen atoms are modeled as united-atoms, which totally reduces the number of atoms of one lipid molecule from 134 for an all-atom model to 52 for the united atom model. The partial charges of the head group are adapted from previous studies [16,17]. The initial configuration of the lipids in their liquid-crystalline phase was equilibrated [17] together with 10 951 TIP3P water molecules [18]. The Particle–Mesh–Ewald method was used for the Coulomb interaction calculation. The total system consists of 40 467 atoms for the melittin case and 40 509 atoms for the case of polyleucine. More details on the preparation of the membrane–water system and the conformational properties of the lipids can be found elsewhere [17].

We used periodic boundary conditions in all three directions. Therefore, it should be noted that in the z-direction the periodic boundary condition leads to a periodic stack of bilayer membranes. At equilibrium the distance between the membranes corresponds to the width of the water ˚. layer which is about 46.7 A The initial structure of melittin was taken from Protein Data Bank (PDB code: 2mlt). This peptide together with six Cl 2 counter ions was solvated with 6509 water molecules. After 500 ps MD simulation at 300 K and 1 bar, the melittin lost part of its a-helicity and became more disordered. Melittin, along with Cl 2 ions and the water molecules within the box defined by the outermost positions of Cl 2 ions, were then transplanted into the water layer of the POPC membrane system by removing almost the same size of box of water molecules. The concentration of melittin corresponds approximately to 10 mM. The initial conformation of Leu25 was taken to be similar to that of melittin, where the backbone structure of Leu25 was taken from that of melittin, and the side chains of melittin were replaced by that of leucine. Since the new polypeptide was totally uncharged at all, the six Cl 2 ions were removed from the model system. Molecular dynamic simulations were performed by the SANDER program in amber [19]. The simulations were performed at T 300 K in the NPT ensemble. The lateral pressure and the perpendicular pressure were P 1 bar. Since explicit water molecules were included in the simulation as solvent, no distance-dependent dielectric constant was used. The atomic coordinates were saved every 1 ps and the atomic velocities were saved every 10 ps to reduce the cost for the need to rerun some part of the simulation. It takes about 0.1 h for 1 ps run on the CRAY T3E using 32 PEs. Since polyleucine is totally uncharged, the partial charges and other force field parameter modifications were expected to be unnecessary. Similar treatment was reported from a translocation study of a shorter polyleucine peptide across the hexane–water interface [20,21]. On the other hand, the positive charged residues of melittin on the N-terminal end, GLY and LYS, would be likely to be deprotonated upon entering the oily phase. But since melittin was still in the aqueous phase during the whole simulation, the change of the protonation state was not expected.

3. Results and discussions 3.1. Adhesion versus insertion

Fig. 2. Density profiles of melittin, water, and lipid heads.

The locations z(t) of the two proteins with respect to the membrane surface as a function of time are presented in Fig. 1. Since the protein is trapped between two adjacent surfaces of a stack of membranes, we have located our fixed coordinate system at the midplane between these two surfaces. The locations of the upper and the lower


99

Fig. 3. Snapshot of melittin. The picture was prepared by using MolScript v2.1 [30].

membrane surfaces are characterized by the average z-coordinate of the phosphate groups, PO4, of the lipid heads of the POPC lipids. According to Fig. 1 the average distance ˚ . This distance is sufficiently between the surfaces is < 47 A large to accommodate the proteins which may adopt a maxi˚ . The locations mum length in the helical state of about 32 A of the two proteins are characterized by the z-components of their centers of mass. From our results of z(t) obtained during MD simulations up to 1 ns, we found no indications towards the onset of insertion into the membrane. Even strong perturbations of the membrane surface were not observed. Rather, the center-of-mass positions of the proteins fluctuate around the midplane of the water region. Direct contact of the molecules with the membrane is very weak, which can be seen from the average density profile of melittin as shown in Fig. 2 where the overlap between the profiles of melittin and the PO4 groups of the lipid heads is very small. Although melittin is located in proximity to only one membrane surface rather than fluctuating between both sides, it is not conclusive whether this indicates an affinity or even adhesion to one of the membrane surfaces. It is equally conceivable that this is an effect from the limited simulation time; for much longer time one could expect a shift or broadening of the density profile. In summary, it should be noted that the melittin is not buried in the interface as expected from experiments [22,23]. There it was estimated that about 40% of the melittin surface is embedded in the hydrophobic part of planar POPC bilayers. In addition, our findings are in contrast to previous Monte Carlo studies on the insertion process of amphiphatic

proteins [4,5,9]. These studies have indicated that melittin at very high dilution must be expected to insert into the hydrophobic core of the membrane, basically driven by the hydrophobic effect, as predicted by the two-state model [1,2]. The main reason for the discrepancy is basically due to the concept of a two-state phenomena which neglects the kinetics of the translocation process. The kinetics, however, is determined by the subtleties of the interfacial barrier near the lipid heads which separate the hydrophobic core of the bilayer from the aqueous medium. It is noted that our result shows different behavior of polyleucine at such a membrane–water interface from that of a previous study for polyleucine at the hexane–water

Fig. 4. The helicity h(t) of melittin and polyleucine as a function of time t.

100


Fig. 5. The bending V (t) of melittin and polyleucine as a function of time t.

interface [20,21]. It was reported that from the beginning polyleucine moved rapidly towards the hexane–water inter˚ /ns. It is very clear from our simulaface at a speed of 11 A tion that the head groups of phospholipids form a very different interface from the hexane–water interface. During the course of our simulation, the hydrophobic peptide has not yet recognized the preferred hydrophobic phase in the acyl chain region.

3.2. Conformational properties Melittin and Leu25 are comparably short proteins which would serve perfectly as single transmembrane-spanning ahelices. In fact, melittin is expected to have a predominant helical structure with a kink approximately at the position 14 caused by proline which do not form hydrogen bonds [24]. In the aqueous medium this structure is not changed significantly. Since melittin is fairly short, it cannot fold properly in order to exhibit a compact tertiary structure which could shield the hydrophobic side chains from

contact with water. A totally coiled structure in water is not expected. The conformation of melittin is remarkably stable over the whole time scale of 1 ns. One snapshot is presented in Fig. 3. The snapshots indicate that melittin has a triangular or open hairpin-like conformation where the hairpin is formed by two a-helices separated by a kink induced by Pro14. The opening angle V < 45 between the two helical axes of the hairpin is remarkably stable over the whole period of simulation and is shown in Fig. 5. The a-helicity of the peptide, h(t), is determined by the O(i) to H–Ni 1 4 distances. According to the amber ‘91 force-field hydrogen bond parameters for these two atom types, we found it is quite adequate to define hydrogen ˚ . Thus a hydrogen bond between resibond length as 2.5 A dues i and i 1 4 is formed when the O(i) to H–Ni 1 4 ˚ . Since there are 26 distance is less than or equal to 2.5 A residues in melittin, a perfect a-helix of melittin could form 22 hydrogen bonds, therefore we can define the total ahelicity h of melittin as the number of hydrogen bonds in this peptide divided by 22. The helicity of melittin as a function of time, h(t), is presented in Fig. 4. It is found that the average helicity of melittin is about khl < 0:75 and does not change significantly as a function of time. Similar observations as for melittin are found for Leu25. Starting with a disordered conformation, this polypeptide assumes after 400 ps an a-helix with a high degree of order. The helicity as function of time is shown in Fig. 4. The bending angles of melittin and polyleucine as a function of time are shown in Fig. 5 and the snapshot after 700 ps is depicted in Fig. 6. The conformational transition from a disordered to a perfect helical structure can also be inferred from the opening angle V (Fig. 5) which is defined by the angle between the two helical axes pointing from the middle of the chain to both ends. After 500 ps the molecule is stretched exhibiting an opening angle close to 1808. The interesting observation is that

Fig. 6. Snapshot of Leu25.


101

A similar procedure can be applied to Leu25. But in this case u is defined by the angle between membrane surface normal and the helical axis. From our data the average angle is < 908 which indicates a parallel orientation to the membrane surface. 4. Remarks and conclusions

Fig. 7. The orientation u (t) of melittin and polyleucine with respect to the membrane normal as a function of time t.

although Leu25 has only hydrophobic side chains, its helical structure is not very much perturbed. This perturbation could be induced by the competition of forming hydrogen bonds either internally or with the surrounding water molecules which could lead to a less rigid helical structure and eventually lead for a much longer helix to a coiled structure. 3.3. Orientation The orientation of melittin at lipid bilayers is controversial. From IR spectroscopy and CD studies [25,26] it was concluded that in dry or partially hydrated (97% relative humidity) bilayers the a-helical portion of melittin is preferentially oriented parallel to the acyl chains. Similarly from polarized ATR-IR spectroscopy of dry phospholipid multibilayers [27]. Magnetic resonance experiments [28,29] indicate a location of melittin on the membrane surface with only the hydrophobic residues buries in the lipid bilayer. It is unclear whether the discrepancy of the different reports on the orientation of melittin interacting with membranes originates from the techniques or from the different type of model membrane preparation which was used for its determination. The orientation of the melittin’s hairpin and the orientation of the helical axis of Leu25 with respect to the surface can be estimated from our MD results. In the case of melittin we define two vectors, aN and aC, for the two branches of the hairpin pointing from residue Pro14 to the N terminus and the C terminus, respectively. The orientational vector is defined as d aN 1 aC =uaN 1 aC u:

1

The orientaional angle u between d and the surface normal is shown in Fig. 7 as a function of time. From the data shown in Fig. 7 it can be concluded that the hairpin has a stable orientation with respect to the membrane surface. The orientation fluctuates around kul < 40 ^ 108.

It should be noted that the validity of our present findings relies heavily, amongst other reasons, on the adequacies of our applied force fields. But this is the current situation and we are looking forward to seeing whether the present force fields have to be modified in the future. Of course, in this situation it would be a great achievement if the reported simulations could be extended in a more systematic way by varying some parameters of the force fields or some external conditions, such as type of lipid head, applied pressure, temperature, among others, in order to elucidate the conditions under which protein insertion takes place. This, however, is a long-term research program because several 10 4 atoms with long range interactions needs a considerable amount of computer time, even on a CRAY T3E. But it is very likely that the future will see such a kind of systematic approach to the lipid–protein interaction using molecular dynamics simulations. Acknowledgements It is a pleasure for us to dedicate the present contribution to Ueli Suter, who has made so many important contributions to polymer science, in particular, in the area of simulations of macromolecules. J.L. is supported by the doctoral fellowship program of the Forschungszentrum Ju¨lich (www.fz-juelich.de/mod/biophys). References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18]

Engelman DA, Steitz TA. Cell 1981;23:411. Jacobs RE, White SH. Biochemistry 1989;28:3421. Jahnig F. Proc Natl Acad Sci USA 1983;80:3691. Milik M, Skolnick J. Proteins Struc Funct Gen 1993;15:10. Milik M, Skolnick J. Biophys J 1995;69:1382. He K, Ludtke S, Heller W, Huang HW. Biophys J 1996;71:2669. Baumgaertner A, Skolnick J. Phys Rev Lett 1995;74:2142. Baumgaertner A. J Chem Phys 1995;103:10 669. Baumgaertner A. Biophys J 1996;71:1248. Simon SM, Peskin CS, Oster G. Proc Natl Acad Sci USA 1992;89:3770. Park PJ, Sung W. J Chem Phys 1998;108:3013. Roseman MA. J Mol Biol 1988;200:513. Eisenberg D, McLachlan AD. Nature 1986;319:199. White SH, Wimley WC. Curr Opin Struct Biol 1994;4:79. Dempsey CE. Biochim Biophys Acta 1990;1031:143. Essmann U, Perera L, Berkowitz ML, Darden T, Lee H, Pedersen LG. J Chem Phys 1995;103:8577–93. Lin J-H, Baumgaertner A. J Mol Liq 1999, in press. Jorgensen WL, Chandrasekhar J, Madura JD, Impey RW, Klein ML. J. Chem. Phys. 1983;79:926–35.

102


[19] Case DA, Pearlman DA, Caldwell JW, Ross WS, Cheatham WTE III, Simmerling, CL, Darden TA, Merz KM, Stanton, RV, Cheng AL, Vicent JJ, Crowley M, Ferguson DM, Radmer RJ, Seibel GL, Singh UC, Weiner PK, Kollman PA. amber 5. University of California, San Francisco, 1997. [20] Chipot C, Pohorille A. J Mol Struct (Theochem) 1997;398–9. [21] Chipot C, Pohorille A. J Am Chem Soc 1998;120:11912–24. [22] Kuchinka E, Seelig J. Biochemistry 1989;2:4216. [23] Beshiaschvili G, Seelig J. Biochemistry 1990;92:52.

[24] [25] [26] [27] [28] [29]

von Heijne G. J Mol Biol 1991;218:499. Vogel H. FEBS Lett 1981;134:37. Vogel H. Biochemistry 1987;26:4652. Frey S, Tamm LK. Biophys J 1991;60:922. Stanislawski B, Ruterjans H. Eur Biophys J 1987;15:1. Altenbach C, Froncisz W, Hyde F, Hubbell JW. Biophys J 1989;56: 1183. [30] Kraulis PJ. J Appl Cryst 1991;24:946.