10146
J. Phys. Chem. B 2007, 111, 10146-10154
Glycolipid Membranes through Atomistic Simulations: Effect of Glucose and Galactose Head Groups on Lipid Bilayer Properties Tomasz Ro´ g,†,‡ Ilpo Vattulainen,§,|,⊥ Alex Bunker,# and Mikko Karttunen*,& Biophysics and Statistical Mechanics Group, Laboratory of Computational Engineering, Helsinki UniVersity of Technology, Espoo, Finland, Department of Biophysics, Faculty of Biotechnology, Jagiellonian UniVersity, Krako´ w, Poland, Institute of Physics, Tampere UniVersity of Technology, Tampere, Finland, Laboratory of Physics and Helsinki Institute of Physics, Helsinki UniVersity of Technology, Espoo, Finland, MEMPHYS-Center for Biomembrane Physics, UniVersity of Southern Denmark, Odense, Denmark, Drug DiscoVery and DeVelopment Technology Center, Faculty of Pharmacy, UniVersity of Helsinki, Helsinki, Finland, and Department of Applied Mathematics, The UniVersity of Western Ontario, London, Ontario, Canada ReceiVed: April 22, 2007; In Final Form: June 8, 2007
Though glycolipids are involved in a multitude of cellular functions, the understanding of their atom-scale properties in lipid membranes has remained very limited due to the lack of atomistic simulations. In this work, we employ extensive simulations to characterize one-component membranes comprised of glycoglycerolipids, focusing on two common glyco head groups, namely glucose and galactose. The properties of these two glycoglycerolipid bilayers are compared in a systematic manner with membranes consisting of phosphatidylcholine (PC) or phosphatidylethanolamine (PE) lipids, whose structures aside from the head group are identical with those of the two glycolipids. We find that the glycolipid systems are characterized by a substantial number of hydrogen bonds in the head group region, leading to membrane packing that is stronger than in a PC but less significant than that in a PE bilayer. The role played by the glyco head group is especially evident in the electrostatic membrane potential, which is particularly large in the glycolipid membranes. For the same reason, the interfacial forces near glycolipid bilayers are significantly different from those found in PC and PE bilayers, affecting, e.g., the ordering of water close to the membrane. These effects are particularly important for the case of galactose, an important component in thylacoids.
1. Introduction Glycolipids are a class of lipids that comprises a very large number of different molecular structures. For example, a subset of glycolipids, known as glycosphingolipids, can be built from about 300 different possible oligosaccharide head groups connected with 60 different hydrophobic moieties.1 Glycolipids are commonly divided into subclasses by the type of organism in which they are found. Typically, glycolipids found in animals belong to the glycosphingolipids2 while those found in plants are glycoglycerolipids.3 The latter have also been observed to appear in certain bacteria.4 The most common glycoglycerolipid head groups are glucose and galactose,3 whereas the most characteristic bacterial glycolipid is lipid A.4 Although glycolipids are usually only expressed in low concentrations in the outer leaflet of animal cell membranes, in certain highly specialized cells, such as those which make up the epithelia, neurons, and myeline, their molar concentration may be as high * Address correspondence to this author. Web: www.softsimu.org. E-mail:
[email protected]. † Laboratory of Computational Engineering, Helsinki University of Technology. ‡ Jagiellonian University. § Tampere University of Technology. | Laboratory of Physics and Helsinki Institute of Physics, Helsinki University of Technology. ⊥ University of Southern Denmark. # University of Helsinki. & The University of Western Ontario.
as 30%.1,5 In plant cells glycolipids are often the dominant class of lipids present, comprising as much as 50% of all lipid molecules in a cell; for example, they are particularly heavily expressed in the photosynthetic membranes.3 It is also notable that the outer membrane of Gram-negative bacteria is predominantly composed of oligosaccharides.4 Glycolipids play a major role in a variety of cellular functions, such as cell-cell recognition, cell adhesion, signal transduction, and protein sorting.6 Many of these functions are related to rafts7shighly ordered, nanosized lipid domains comprised of saturated sphingolipids, glycolipids, and cholesterol. The majority of cellular glycosphingolipids are found in rafts.1,6 The interest in the mechanics, structure, and function of rafts has been greatly increased by recent findings, which suggest that rafts provide specific environments for membrane proteins.7 In addition to rafts, interest toward glycolipids has increased due to their importance in thylakoids. Plant glyceroglycolipids are the major structural element of thylakoids, the dominant component of photosynthetic membranes (up to 75%3). Their function is to ensure the proper curvature of thylakoids8 and the stabilization of large photosynthetic protein complexes.9 They also play a role in maintaining the transmembrane ionic gradients.2 From a biophysical point of view, the main difference between glycolipids and phosphatidylcholines (PCs)sa very common lipid class in essentially all cellular membranessis the ability of the sugar head groups to participate in hydrogen
10.1021/jp0730895 CCC: $37.00 © 2007 American Chemical Society Published on Web 08/03/2007
Glycolipid Membranes through Atomistic Simulations bonds as both donors and acceptors,10 while the PC head group can only act as an acceptor. As a result, the main phase transition temperature (Tm) of glycoglycerolipids with a monosaccharide head group is higher than that of the corresponding PCs. In addition, the surface area of glycolipids in the liquid-disordered (fluid) phase is lower than that of PCs,11 supporting the idea that glycolipids could promote the stabilization of lipid membranes. Experimental NMR studies of glycosphingolipids12 support this view, since they have shown that the conformational order along the glycosphingolipid chains is slightly higher than that for corresponding chains in phospholipids. For long oligosaccharides, phase behavior becomes complicated and strongly dependent on sugar composition (for a review, see refs 2 and 13). Monogalacto-diacyl-glycerols extracted from photosynthetic membranes are also known to form various nonlamellar phases such as the hexagonal-II or the cubic phase.14 Studies on the effect of sugar head groups on acyl chain order have thus far been limited to the case of galacto-cerebroside in mixtures with 1-stearoyl-2-oleoyl-phosphatidylcholine (SOPC).15 In addition to their role in biological membranes, glycolipids also have a role to play as a component in liposome-based nanotechnological devices designed to achieve targeted drug delivery and gene therapy. These devices involve surrounding the gene or protein to be delivered with a cationic lipid bilayer (in the case of gene therapy) or neutral or negatively charged lipids in the case of drug delivery, tethered PEG molecules for avoiding detection by the immune system, and targeting ligands to ensure delivery to a specific cell type. Glycolipids have been included as targeting ligands, for example, the glycerosphingolipid sulfatide, known to act as a cellular receptor,16 has been shown to be an effective targeting mechanism for cancer therapy, as it acts as a receptor for extracellular matrix glycoproteins, highly upregulated in many varieties of tumor cells.17 In addition, sulfatide has been found to greatly enhance the stability of liposomes into which it has been incorporated. This is thought to be due to the hydration of the negatively charged sulfatide head group.18 Other glycolipids used have been galactose for targeting galactose receptors in the liver, for the treatment of liver cancer,19 and mannose for the targeting of macrophages.20 While atom-scale molecular dynamics (MD) simulations have been successfully carried out for a variety of one- and twocomponent lipid systems comprised of, e.g., PCs, sphingolipids, and sterols (see, for example, refs 21-24 and references therein), and recently also for many-component membrane systems related to lipid rafts,25-28 there have so far been very few studies of lipid bilayers incorporating sugar molecules. These few have focused on bilayer properties and parametrization,29,30 the interactions of trehalose with PC bilayers,31,32 bilayers composed of bacterial lipopolysaccharides,32,33 galactosylceramides,34 and gangliosides.35,36 These studies have provided substantial insight into the ways in which sugar molecules interact within membranes and how they are able to stabilize them. In the meantime, it is rather surprising that, to our knowledge, there are no previous simulation studies of glycoglycerolipids. In the present work, we therefore focus on glycolipid membranes by considering the effects of four different head groupssglucose, galactose, PC, and phosphatidylethanolamine (PE)son lipid bilayer properties. The glucose and galactose head groups are the essence of the glycolipids studied here. To better understand the properties that result from their specific nature, we compare them systematically to the results obtained for the PC and PE lipids whose structures except for the head group region are identical with those of the two glycolipids.
J. Phys. Chem. B, Vol. 111, No. 34, 2007 10147 2. Methods 2.1. System Description and Parameters. We have performed atomic-scale molecular dynamics simulations for four different single-component membrane systems, each consisting of a total of 128 lipids. The first bilayer was composed of 1,2di-O-palmitoyl-3-O-β-D-glucosyl-sn-glycerol (DP-GLUC) molecules, the second of 1,2-di-O-palmitoyl-3-O-β-D-galactosylsn-glycerol (DP-GALA) lipids, the third of dipalmitoylphosphatidyl-choline (DPPC) molecules, and the fourth of dipalmitoyl-phosphatidyl-ethanolamine (DPPE) lipids. Figure 1 shows the structure and the numbering of atoms in DP-GLUC, DP-GALA, DPPC, and DPPE molecules. Notice that the hydrocarbon chain and the glycerol regions of the four molecules are identical and only the head group region is varied from one molecule to another. This allows us to study the effects of the glyco head groups on membrane properties in a systematic manner. All four bilayers were hydrated with approximately 3500 water molecules. The initial structures of all membranes were obtained by arranging the lipid molecules in a regular array in the bilayer (x, y) plane with an initial surface area of 0.64 nm2 per lipid molecule. Prior to the actual MD simulations, the steepest-descent algorithm was used to minimize the energy of the initial structure.37,38 The simulations were performed by using the GROMACS software package.39,40 The MD simulations of all membrane systems were carried out over 50 ns. The first 10 ns were considered as an equilibration period,41 as the results for the average area per lipid indicated that it settled to its equilibrium value after about 4-6 ns. Consequently, only the last 40 ns of the trajectory were analyzed. Figure 2a depicts a snapshot of an equilibrated DP-GLUC bilayer, and Figure 2b shows the space filling models of glucose, PC, and PE head groups. To parametrize all lipid molecules we used the all-atom OPLS force field42 extended for carbohydrate simulations.29 Partial charges are part of the OPLS parametrization. Charge groups are small, neutral units originally derived in the parametrization.29 As glucose and galactose differ only in chirality, we used the same topology for both of them. Partial charges on the PC head groups were taken from Takaoka et al.43 and on PE from Murzyn and Pasekiewicz-Gierula.44 Both sets of charges were derived in compliance with the OPLS methodology. In Figure 1 charge groups are marked on the chemical structures of the lipids used in this study. In the hydrocarbon tails each methylene or methyl group was treated as a separate charge group. For water, we employed the TIP3P model which is compatible with the OPLS parametrization.45 Periodic boundary conditions with the usual minimum image convention were used in all three directions. The LINCS algorithm46 was used to preserve hydrogen covalent bond lengths. The time step was set to 2 fs and the simulations were carried out at constant pressure (1 bar) and temperature (343 K). The simulation temperature was chosen such that it is above the main phase transition temperature of DP-GLUC and DP-GALA11 (about 339 K), and also above the main transition temperatures of DPPC (314 K) and DPPE (338 K). The temperature and pressure were controlled by using the Berendsen method47 with relaxation times set to 0.4 and 1.0 ps, respectively. The temperatures of the solute and solvent were controlled independently. For pressure, we used a semi-isotropic control. The Lennard-Jones interactions were cut off at 1.0 nm, and for the electrostatic interactions we employed the particlemesh Ewald method48 with a real space cutoff of 1.0 nm, β-spline interpolation (order of 5), and direct sum tolerance of
10148 J. Phys. Chem. B, Vol. 111, No. 34, 2007
Ro´g et al.
Figure 1. Molecular structures of (a) DP-GLUC, (b) DPPC, and (c) DPPE molecules with numbering of atoms.
10-6. The simulation protocol used in this study has been successfully applied in various molecular dynamics simulation studies of lipid bilayers.24,30,38,41,49 2.2. Analysis. In the following discussion, we consider various quantities determined from the simulation data. Surface area/PC was calculated by dividing the total area of the membrane by the number (64) of PC molecules in a single leaflet. Membrane thickness was determined from mass density profiles by considering the points where the mass densities of lipids and water merge.38 It must be noted that a quantitative comparison with experimental data is typically not straightforward.50 The molecular order parameter (Smol) described in detail elsewhere51 provides essentially the same information as the commonly studied NMR order parameter SCD.52 For the present saturated chains of DPPC, Smol ) 2|SCD|. To calculate the tilt angles for the acyl chains of DPPC, we averaged over seg-
mental vectors g4 the nth segmental vector links carbon atoms n and n + 1 in the acyl chain to obtain the average segmental vector. The tilt angle for a given acyl chain is then given by 〈arccos(xcos2θ)〉, where θ is the angle between the bilayer normal and the average segmental vector.53 In averaging conformational quantities in terms of (gauche) and (trans) states, only the torsion angles 4-16 were taken into account.51 To analyze hydrogen bonding, water bridging, and charge pairing, we employed the same geometrical definitions as in our previous papers.54-56 Charge pairing, which essentially describes the electrostatic interaction between a positively charged molecular moiety such as a methyl group in PC choline and a negatively charged one such as an oxygen atom in the sterol OH group, complements our studies for atomic-scale interaction mechanisms and is most useful in describing interactions in the head group region.
Glycolipid Membranes through Atomistic Simulations
Figure 2. Snapshot of equilibrated DP-Gluc bilayer after 50 ns of MD simulation (a); space filling models of glucose, PE, and PC head groups (b); and structures of two DP-GLUC molecules connected through direct hydrogen bond (c) and water bridge (d).
3. Results 3.1. Membrane Dimensions. The calculated surface areas are given in Table 1. For the present membrane systems, they can be compared with experimental findings in all the cases. For DP-GLUC and DP-GALA the surface area has been measured over a large range of temperatures by means of X-ray scattering. For DP-GLUC at 343 K it has been found to be ∼0.61 nm2, and for DP-GALA ∼0.60 nm2.11 These values are consistent with our simulation results, ∼0.60 nm2. For DPPC and DPPE, experimental values of surface areas have been determined from the NMR SCD, close to the value of 0.58 nm2 obtained from our simulation. For DPPC, extrapolating the experimental results to the temperature 343 K (values are given only for 338 and 353 K) gives a value of 0.69 nm2, in close
J. Phys. Chem. B, Vol. 111, No. 34, 2007 10149 agreement with our simulation result of 0.70 nm2. We conclude that in all membrane systems studied here, the simulation results are consistent with the available experimental data. The results also highlight the stabilizing effect of glyco head groups with respect to PCs. Changes in surface area are often associated with changes in membrane thickness. The values for membrane thickness (see Table 1) obtained for the glycolipid and DPPE membranes are similar, while the value for DPPC is much lower, consistent with the results for surface area. To visualize differences in membrane thicknesses we also plotted partial density profiles for the lipids in all bilayers (Figure 3a). To show the position of DP-GLUC we made a separate plot of the partial density profiles of the head group, glycerol moiety, both acyl chains, and water (Figure 3b, analogous profiles for DP-GALA do not differ; data not shown). On the density profile of DPPE, asymmetry between the two leaflets is visible. This results from local deformation (invagination) of the surface in the head group region of the lower leaflet. This deformation is likely related to the ability of this combination of lipids to form a nonlamellar phase, and our hypothesis is that this could possibly be an inverted micelle phase if the systems were large enough. The deformation is present throughout most of the simulation time. To provide a comparison with experiments, we also estimated the volume and thickness of the DP-GLUC system. The membrane thickness obtained in this study for DP-GLUC is 4.31 nm. With the thickness (4.31 nm) and the surface area (0.602 nm2), it is possible to estimate the volume of DP-GLUC. That gives 1.297 ( 0.04 nm3. The partial volume measured by means of densitometry11 is about 1.2 nm3, and thus we can conclude that the experimental and computational data are in good agreement. 3.2. Order and Conformation of Acyl Chains. The difference in the structure of head groups leads to a rather profound difference in the ordering of the acyl chains. This is illustrated by the molecular order parameter, Smol, whose profiles along the sn-1 and sn-2 chains of DP-GLUC, DP-GALA, DPPC, and DPPE are shown in Figure 4. For mean values (averages over segments 4-16) of Smol for the sn-1 and sn-2 chains, see Table 1. The results highlight the fact that the glycolipid systems are considerably more ordered than the PC counterpart. Due to the lack of experimental data, direct comparison to NMR data for glycosphingolipids is not possible. A similar trend, however, has been observed between glycosphingolipid and PC bilayers.12 Figure 4 and Table 1 clearly illustrate the major difference between lipids whose head groups are hydrogen bonding (GLUC, GALA, and PE) and the one that does not have a similar capability (PC). Distributions of the tilt angle of the sn-1 and sn-2 chains are shown in Figure 5, and the corresponding average values given in Table 1 support this view. Evidently, DPPC chains are more tilted than chains of PE and glycolipids. As for isomerization and its dependence on membrane composition, we found that the differences between the average numbers of (gauche) states per chain are small and statistically insignificant (Table 1). The average lifetime of (trans) conformations in acyl chains (Table 1) is longest for DPPE, intermediate for glycolipids, and shortest for DPPC indicating fastest dynamics along the acyl chains for the last system. The differences in lifetimes are small, however, and thus not fully conclusive. 3.3. Head Group Orientation and Dynamics. To describe the orientation of the sugar head groups we determined the orientation of vectors between the atoms O4-C12 in DP-GLUC and DP-GALA relative to the bilayer normal. Distribution of the angle between this vector and bilayer normal is shown in
10150 J. Phys. Chem. B, Vol. 111, No. 34, 2007
Ro´g et al.
TABLE 1: Ordering and Condensing Effects of Sterolsa Smol tilt (deg) no. of gauche defects lifetime (ps) area/lipid (nm2) thickness (nm)
chain
DP-GL 1
DP-Gal 1
DPPE
DPPC
sn-1 sn-2 sn-1 sn-2 sn-1 sn-2 sn-1 sn-2
0.35 ( 0.01 0.38 ( 0.01 20.1 ( 0.1 19.3 ( 0.1 4.19 ( 0.01 4.26 ( 0.01 57 ( 2 58 ( 2 0.602 ( 0.01 4.31 ( 0.02
0.36 ( 0.01 0.38 ( 0.01 19.6 ( 0.1 19.4 ( 0.1 4.17 ( 0.01 4.23 ( 0.01 57 ( 2 58 ( 2 0.598 ( 0.01 4.27 ( 0.02
0.40 ( 0.01 0.42 ( 0.01 18.8 ( 0.1 19.2 ( 0.1 4.00 ( 0.01 4.03 ( 0.01 60 ( 2 61 ( 2 0.580 ( 0.01 4.27 ( 0.02
0.24 ( 0.01 0.28 ( 0.01 23.1 ( 0.1 22.2 ( 0.1 4.49 ( 0.01 4.49 ( 0.01 54 ( 2 55 ( 2 0.700 ( 0.01 3.85 ( 0.02
a Average values of the molecular order parameter, Smol, chain tilt angle, number of gauche states per acyl chain, and lifetimes of trans conformations. All results are given separately for the sn-1 and sn-3 chains. Also given here are the average surface area per lipid and the membrane thicknesses.
Figure 3. Partial density profiles along the bilayer normal. (a) Shown here are all bilayer atoms in DP-GLUC (black line), DP-GALA (dashed black line), DPPE (gray line), and DPPC (dashed gray line) bilayers. (b) In a similar manner, data are shown in the DP-GLUC bilayer for water (dotted black line), lipid (gray line), head group (black line), glycerol backbone (dashed black line), sn-1 chain (dashed gray line), and sn-2 chain (thin black line). The coordinate z ) 0 corresponds to the membrane center.
Figure 5. Distribution of tilt angles of (a) sn-1, and (b) sn-2 chains in DP-GLUC (black line), DP-GALA (dashed black line), DPPE (gray line), and DPPC (dashed gray line) bilayers.
Figure 6. Distribution of angles between the O4-C12 vector and bilayer normal in DP-GLUC (black line) and DP-GALA (dashed black line) bilayers.
Figure 4. Profiles of the molecular order parameter Smol calculated for (a) sn-1 and (b) sn-2 chains in DP-GLUC (black line), DP-GALA (dashed black line), DPPE (gray line), and DPPC (dashed gray line) bilayers.
Figure 6. We see that the orientations of both head groups are similar. The head group orientation is mostly determined by the torsion angles around C1-O1, O1-C11, and C11-C12 bonds, thus we also examined the conformations of these angles. Consistent with the orientation of sugar molecules, no differences were found between DP-GLUC and DP-GALA. Rota-
Figure 7. Rotational autocorrelation function of the O4-C12 vector in DP-GLUC (black line) and DP-GALA (dashed black line), and of the P-N vector in DPPE (gray line) and DPPC (dashed gray line) bilayers.
tional autocorrelation functions (RACFs)22 of the vector O4C12 in DP-GLUC and DP-GALA and the P-N vector in DPPE and DPPC are shown in Figure 7. The fastest rotation was observed for DPPC, an intermediate rate was observed for both sugar head groups, and the slowest rate was observed for DPPE.
Glycolipid Membranes through Atomistic Simulations
J. Phys. Chem. B, Vol. 111, No. 34, 2007 10151 TABLE 2: Interactions in the Membrane/Water Interface: Lipid-Water, Lipid-Lipid, Hydrogen Bonds, Water Bridges, and Charge Pairs (Errors
