Article Information

• PDF (536 kb)
• Full Text with Large Figures
• Cited by in Scopus (2)

PubMed

• Articles by Jirasak Wong-ekkabut
• Articles by Luca Monticelli

Related Articles

• Glutathione Peroxidase 4 Senses and Translates Oxidative Str...

  Glutathione Peroxidase 4 Senses and Translates Oxidative Stress into 12/15-Lipoxygenase Dependent- and AIF-Mediated Cell Death Cell Metabolism, Volume 8, Issue 3, 3 September 2008, Pages 237-248 Alexander Seiler, Manuela Schneider, Heidi Förster, Stephan Roth, Eva K. Wirth, Carsten Culmsee, Nikolaus Plesnila, Elisabeth Kremmer, Olof Rådmark, Wolfgang Wurst, Georg W. Bornkamm, Ulrich Schweizer and Marcus Conrad Summary Oxidative stress in conjunction with glutathione depletion has been linked with various acute and chronic degenerative disorders, yet the molecular mechanisms have remained unclear. In contrast to the belief that oxygen radicals are detrimental to cells and tissues by unspecific oxidation of essential biomolecules, we now demonstrate that oxidative stress is sensed and transduced by glutathione peroxidase 4 (GPx4) into a-yet-unrecognized cell-death pathway. Inducible GPx4 inactivation in mice and cells revealed 12/15-lipoxygenase-derived lipid peroxidation as specific downstream event, triggering apoptosis-inducing factor (AIF)-mediated cell death. Cell death could be entirely prevented either by α-tocopherol (α-Toc), 12/15-lipoxygenase inhibitors, or siRNA-mediated AIF silencing. Accordingly, -deficient cells were highly resistant to glutathione depletion. Neuron-specific GPx4 depletion caused neurodegeneration in vivo and ex vivo, highlighting the importance of this pathway in neuronal cells. Since oxidative stress is common in the etiology of many human disorders, the identified pathway reveals promising targets for future therapies. Summary | Full Text | PDF (1825 kb)

• Spontaneous and thermoinduced photon emission: new methods t...

  Spontaneous and thermoinduced photon emission: new methods to detect and quantify oxidative stress in plants Trends in Plant Science, Volume 8, Issue 9, 1 September 2003, Pages 409-413 Michel Havaux Abstract Peroxidation of polyunsaturated fatty acids is one of the main events triggered by oxidative stress in cells. Some lipid peroxidation products are light-emitting species, and their luminescence can be used as an internal marker of oxidative stress. However, this spontaneous chemiluminescence is weak and difficult to measure. Recent studies have shown that an alternative approach that involves measuring thermoluminescence bands at high temperature (in the range 80–150°C) is a simple way of detecting and quantifying lipid peroxidative damage and oxidative stress in plants. Abstract | Full Text | PDF (159 kb)

• Alkyl hydroperoxide reductases: the way out of the oxidative...

  Alkyl hydroperoxide reductases: the way out of the oxidative breakdown of lipids in chloroplasts Trends in Plant Science, Volume 4, Issue 5, 1 May 1999, Pages 166-168 Margarete Baier and Karl-Josef Dietz Full Text | PDF (85 kb)

• …more

Copyright © 2007 The Biophysical Society. All rights reserved.
Biophysical Journal, Volume 93, Issue 12, 4225-4236, 15 December 2007

doi:10.1529/biophysj.107.112565

Membranes

Effect of Lipid Peroxidation on the Properties of Lipid Bilayers: A Molecular Dynamics Study

Jirasak Wong-ekkabut^*, †, Zhitao Xu^*, Wannapong Triampo^†, ‡, I-Ming Tang^†, §, D. Peter Tieleman^* and Luca Monticelli^*, ,  

^* Department of Biological Sciences, University of Calgary, Calgary, Alberta, Canada
^† Department of Physics, Mahidol University, Bangkok, Thailand
^‡ Center for Vector and Vector-Borne Diseases, Mahidol University, Bangkok, Thailand
^§ Center of Nanoscience and Nanotechnology, Mahidol University, Bangkok, Thailand

Address reprint requests to Luca Monticelli, Tel.: 403-220-4039.

Lipid peroxidation alters the physiological functions of cell membranes and plays an important role in cellular membrane damage. Peroxidation is believed to be involved in cellular aging and in various diseases, such as Parkinson’s and Alzheimer’s disease 1,2,3,4,5,6,7,8,9 as well as schizophrenia 10, atherosclerosis 11,12, inflammatory diseases 13, and cardiac ischemia reperfusion injury 14,15. Unsaturated lipids are easily susceptible to peroxidation 16. The effect of both unsaturation and peroxidation on the properties of lipid bilayers has been well characterized experimentally 17,18,19,20,21,22. Still, the exact mechanism of membrane damage by oxidized lipids is unclear. Oxidized lipid tails are more polar and can be shorter in length, due to the presence of aldehyde or hydroperoxide groups 23,24. Lipid peroxidation has been shown to perturb the bilayer structure and modify membrane properties such as membrane fluidity, permeability to different substances, and bilayer thickness. The existence of a direct relationship between lipid peroxidation and membrane leakiness has been suggested 25,26,27,28. Increased membrane permeability caused by oxidation of lipids and membrane proteins can disrupt ion gradients, therefore altering metabolic processes. Lipid peroxidation can influence the permeability of lipid membranes by increasing the dielectric constant of the membrane interior and by increasing the microviscosity, possibly through cross-linking of lipid radicals 23. Focusing on structural and dynamic properties, a decrease in membrane thickness upon oxidation has been observed using x-ray diffraction analysis, along with interdigitation of the terminal methyl segments 22. The effect of peroxidation on lipid dynamics and membrane order is less clear. According to some researchers, peroxidation does not affect the fluidity of the membrane 29 nor the reorientational dynamics of the lipids 18. According to others, membrane fluidity is decreased 30,31,32,33 and the decrease is higher near the double bonds of the bilayer, whereas other regions are less affected 30. Some have reported an increase in the lipid tails order parameter 30,33,34,35,36, others no change 37 or a decrease 18,38. Several reasons can explain the differences in experimental results, including the use of different methodologies to generate peroxides, leading to different (and usually not well-defined) lipid compositions of the membrane 22. Despite the numerous studies on the effect of oxidation on the structure and dynamics of lipid membranes, the relationship between increased membrane permeability, and modifications in the structure and dynamics of lipid bilayers is not clear.

In recent years, computational studies of model membranes proved to be particularly useful in the description of the structure and dynamics of lipid bilayers 39,40 and in the interpretation of experimental results 41. Pure lipid bilayers, including either a single type of lipid 42,43,44 and mixtures of two or more lipids 43,45,46,47,48,49,50, have been investigated using computational methods, as well as mixtures of lipids with proteins 51. Unsaturated lipid bilayers also have been studied using computer simulations 52,53,54,55,56, but the structural consequences of the presence of oxidized lipids have never been investigated using computational methods, to the best of our knowledge.

In this work, we use molecular dynamics simulations to characterize the effect of lipid oxidation on the properties of 1-palmitoyl-2-linoleoyl-sn-glycero-3-phosphatidylcholine (PLPC) lipid bilayers. In particular, we describe the effect of four different products of PLPC peroxidation 57 at five concentrations, ranging from 2.8% to 50%. We focused on four main oxidation products of linoleic acid, including either a hydroperoxide or an aldehyde group: 9-trans, cis-hydroperoxide linoleic acid (9-tc), 13-trans, cis-hydroperoxide linoleic acid (13-tc), 9-oxo-nonanoic acid (9-al), and 12-oxo-9-dodecenoic acid (12-al) (Fig. 1). These oxidized chains replaced the sn-2 linoleate chain in PLPC. The goal of this work is to understand how oxidized lipids change the membrane properties. In particular, we seek to characterize the relationship between the changes in membrane permeability and the modifications of structural and dynamic properties of the lipid bilayer, and to provide a detailed description at the atomistic level of the chemical interactions responsible for the changes in the properties of the membrane. The article is organized as follows. First, we describe the methods used to derive the simulation parameters for the lipids and the simulation methods. Then we describe a number of properties of a PLPC bilayer and how they change upon increasing the concentration of oxidized lipids, and compare simulation results to experimental ones reported in the literature. Finally, we discuss the relationship between permeability, structure, and dynamics of oxidized bilayers.

Display large version of this figure

Figure 1

Products of the oxidation of linoleic acid considered for this work.

A united-atom force field was used for the lipids in all simulations. The parameters for the phosphatidylcholine (PC) headgroup and the lipid tails were taken from previous works on PLPC and DPPC lipids 42,54. The hydroperoxide lipid tails were created by addition of a hydroperoxide group at position C9 or C13 of the linoleate tail and shifting the double bonds, as shown in Figure 1ab. The aldehyde lipid tails were also built starting from linoleic acid (Figure 1cd). The bonded parameters for the O-O and O-H bonds and for the O-O-H angle were taken from previous calculations on hydrogen peroxide 58. The dihedral angle parameters and the partial charges for the peroxide and aldehyde groups were derived using quantum chemistry calculations on 3-hydroperoxy-1-butene and propanal (Figure 2ab), while Lennard-Jones parameters were taken from hydroxide and carbonyl groups already present in the force field. We used the Jaguar software package 59 for all quantum calculations, with the B3LYP method of density functional theory 60,61 and the LACV3P**++ basis set. Partial atomic charges were estimated using natural population analysis 62 and the electrostatic surface potential fitting method with Merz-Kollman atomic radii 63 after the geometry optimization. The results for partial charges are reported in Table 1. For the calculation of bond and angle force constants, we restrained the bond lengths and angles at seven different values, then fitted a harmonic potential functions to the energy profile. For dihedral parameters, dihedral angles were restrained at 36 different values from 0 to 360°, and the standard proper dihedral function was fitted to the potential energy. For all bonded parameters, the Lennard-Jones and electrostatic energy were calculated for different geometries and subtracted from the total energy before fitting. Results for all bonded parameters are reported in Table 2,Table 3.

Display large version of this figure

Figure 2

Lipid fragments used for the calculation of bonded parameters and partial charges for the oxidized lipid tails.

Table 3 Force constants for dihedral angles in the peroxide and aldehyde groups

	Dihedral angles	a*	b†	c*	d†	e*	f†	Reference molecule
	C=C-C-O	2.12	223.90	0	0	3.62	180.50	3-hydroperoxy-1-butene
	C-C-O-O	2.13	334.25	0	0	7.04	8.10	3-hydroperoxy-1-butene
	C-O-O-H	8.46	23.30	6.51	18.40	0	0	3-hydroperoxy-1-butene
	C-C-C=O	0.47	180.00	1.58	180.00	2.67	180.00	propanal

The functional form is the following:

* kJ/(mol×rad2). † Degrees.

The systems were generated starting from the equilibrium structure of a PLPC bilayer containing 72 lipids. We replaced 2, 4, 8, 18, and 36 PLPC lipid molecules with each oxidized lipid, obtaining 20 different bilayers with oxidized lipid concentrations of 2.8%, 5.6%, 11.1%, 25%, and 50%, respectively. The two bilayer leaflets always contained the same number of oxidized lipids. All 21 simulation systems contained 72 lipid (including PLPC and oxidized lipids) and 2880 water molecules. All simulations were carried out with version 3.3.1 of the GROMACS package 64. After energy minimization, molecular dynamics simulations were run for 180ns, and the initial 80ns were considered as an equilibrium period. The integration time step was 2fs. Periodic boundary conditions were applied in all dimensions. A 1.0nm cutoff was employed for the electrostatic and Lennard-Jones interactions and the neighbor list was updated at every time step. The long-range electrostatics was calculated using particle-mesh Ewald 65,66; the real-space interactions were evaluated using a 1.0nm cutoff, and the reciprocal-space interactions were evaluated on a 0.12nm grid with fourth-order B-spline interpolation. The relative error for the Ewald sum in the direct and reciprocal space, controlled in GROMACS by the parameter ewald_rtol, was set to 10⁻⁵. The LINCS algorithm was used to constrain all bond lengths 67. The weak temperature coupling scheme was applied separately to the lipids and water 68, with a temperature of 298K and a time constant of 0.1ps. The semiisotropic pressure was applied 68, with an equilibrium pressure of 1 bar both in the x-y plane and in the z direction (bilayer normal) with a time constant of 4.0ps and a compressibility of 4.5×10⁻⁵bar⁻¹. Molecular graphics were made using VMD 69.

Constraint simulations were used to calculate the potential of mean force (PMF) of water as a function of the distance from the center of the bilayer, the local diffusion coefficient at different depths, and water permeability through the lipid bilayer 70,71. A series of 31 simulations was run with the distance between water and the center of bilayer constrained between 0 and 3.0nm, with 0.1nm spacing. Only the component of the distance along the bilayer normal (z axis) was constrained, while water was completely free to move in the x and y directions. The SHAKE algorithm was used, with a relative tolerance of 10⁻⁵. Two water molecules were constrained at the chosen depths inside the bilayer, at a distance of 3.0nm (along the z axis) from each other. In the first simulation, one water molecule was restrained at 0nm (corresponding to the center of the bilayer) and the second at 3nm (corresponding to the bulk water phase). This setup allows for increased sampling at no computational cost. Each simulation was 15ns long and the forces were calculated as a function of the simulation time. The free energy of water transfer from the bulk phase to various depths in the membrane can be expressed as

(1)

(2)

The first important change in the simulations of all the oxidized lipids is in the conformation of the lipid tails. Snapshots showing the conformation of oxidized and nonoxidized lipids are shown in Fig. 3. The portion of the lipid tail containing oxygen atoms is found, on average, close to the interface region. This is confirmed by the distribution of aldehyde and peroxide oxygen atoms in the bilayer, shown in Fig. 4. For both the aldehyde-containing and the peroxide-containing lipids, the maximum density of oxygen atoms is around the carbonyl group, and the distribution is broader for aldehyde-containing lipids. Together with the conformational change, hydrogen bonding is observed between the oxidized lipid tail and water, carbonyl groups, and phosphate groups. Table 4 shows the average number of hydrogen bonds formed by the hydroperoxide and aldehyde groups with other groups, in each simulation. In all cases, oxidized lipid tails form hydrogen bonds mostly with water molecules. For hydroperoxide-containing lipids, hydrogen bonds with phosphate groups are more probable than with carbonyl groups. The average total number of hydrogen bonds per lipid does not change significantly with the concentration of oxidized lipids. Its average value is 1.00±0.13 for hydroperoxide lipids and 0.48±0.05 for aldehyde lipids. This highlights a correlation between oxygen density distribution and hydrogen bonding: hydroperoxide lipids have higher propensity to form hydrogen bonds with water and narrower density distributions. These findings corroborate the model initially proposed by van Kuijk et al. 74, suggesting that the hydroperoxide moieties reside in the proximity of the lipid headgroup region, because of their hydrophilic character.

Display large version of this figure

Figure 3

Snapshots of a single PLPC and 13-tc taken at 5ns intervals. Lipids are colored by atom type: nitrogen is blue, carbon is cyan, oxygen is red, and phosphorus is tan. Molecules are oriented along the z axis and superimposition was done on the phosphorus and oxygen atoms.

Display large version of this figure

Figure 4

Electron density profiles for the oxygen atoms of the oxidized acyl chains and the phosphorus atoms of the headgroup in PLPC bilayers containing 11.1% of oxidized lipids.

The presence of hydrogen-bonding interactions involving the lipid tails affects most of the properties of the lipid bilayer. Figure 5A shows the electron density profile calculated from our simulations of pure PLPC and for bilayers containing 50% concentration of each oxidized lipid. The total density at the center of the bilayer is increased in the presence of oxidation, and the maxima are shifted toward the center. The increase of the density at the center of the bilayer corresponds to partial interdigitation of the phospholipids acyl-chain terminal methyl segment when the thickness of bilayer decreases. While experimental data on PLPC are not available, the change in the electron density profile upon peroxidation has been characterized experimentally for dilinoleoyl phosphatidylcholine (DLPC) bilayers, as shown in Figure 5B (reproduced from 22). Simulation results on PLPC compare favorably with the experimental ones on DLPC, showing a decrease in the bilayer thickness and a higher density in the center.

Display large version of this figure

Figure 5

(A) Total electron density in simulations of PLPC bilayers including two oxidation products, 12-al and 9-tc, at 50% concentration. (B) Relative electron density in x-ray diffraction experiments on DLPC and peroxidized products (reproduced from 22).

We calculated the average area per lipid and bilayer thickness in the 21 simulated systems (Table 5), and compared the results to previous simulations and experimental data. The thickness of the bilayer was calculated from the simulations as the average distance between phosphate groups in the two leaflets, computed from the total electron density profile. Errors were estimated using a block averaging procedure described by Hess 75. Fig. 6 shows the area per lipid molecule and the bilayer thickness at different concentrations of each oxidized lipid. For the pure PLPC bilayer, we found an average area per lipid of 0.651±0.015nm² and an average thickness of 3.62±0.01nm. The difference with previous calculations 54,76 and experimental findings 53,77 is within 6% for the area and 3% for the thickness. For all the bilayers containing oxidized lipids, the area increases with increasing concentrations of the oxidized lipids, and the thickness decreases. Visual inspection of the trajectories suggests that the increase in area per lipid and the corresponding decrease in the thickness are related to the preference of the more polar oxidized tails for the interface and the headgroup region. The relationship between area and thickness is not straightforward, since both the length of the oxidized lipid tail and the position of the oxygen in tail have a specific effect on the structural properties of the bilayer. The thickness is generally less when the bilayer contains aldehyde lipids, for which one of the acyl tails is shorter. On the other hand, lipids with the peroxide or aldehyde groups farther away from the carbonyl ester tend to give larger areas. Bilayers containing 13-tc generally have the largest area per lipid, but not the smallest thickness.

Display large version of this figure

Figure 6

The thickness and area per lipid of lipid bilayer, containing various concentrations of each oxidized lipid. Statistical error estimates are less than the size of the symbols for all the simulations (see Table 5), and are therefore omitted for the sake of clarity.

The increase in the area per lipid observed in our simulations is consistent with experimental results by Pradhan et al. 78 showing that peroxidized lipids increase the phospholipids spacing in erythrocyte membranes. Sabatini et al. 79 characterized DPPC monolayers containing oxidized lipids, in particular 9-al (referred to as PoxnoPC, in their study) and the carboxylic acid analog. They found that both oxidized lipids expanded the monolayers, similarly to our results. Interestingly, film expansion was larger with the carboxylic terminal group, the more polar group. They also proposed a model for the arrangement of the sn-2 acyl chains in monolayers that is similar to the model of van Kuijk 74, and consistent with our results.

The degree of ordering of the tail is also influenced by the presence of the oxygen atoms, as shown by the deuterium order parameter. The deuterium order parameter can be measured by NMR, and is defined as

(3)

C double bonds. The order parameter was calculated for all the CD bonds in the sn-1 and sn-2 chains of both PLPC and the oxidized lipids in all the simulations. Figure 7A shows the deuterium order parameter calculated on the sn-2 chain in the simulation of pure PLPC and in the presence of 11% oxidized lipids. Standard errors were calculated using a block averaging procedure. Table 6 reports the average order parameter for the sn-1 and the sn-2 chains of PLPC and each oxidized lipids in all the simulations. As expected, both in PLPC and in the oxidized lipids the unsaturated tail has lower order parameters compared to the saturated one. In general, the order parameter of the acyl chains for all the lipids decreases with increasing concentration of oxidized lipids. Figure 7BC, show the deuterium order parameter of the sn-1 chain of PLPC in the presence of 9-tc and 9-al at different concentrations. The disordering effect induced by the presence of the oxidized lipid tail is stronger for aldehyde lipids than for hydroperoxide ones (see also Table 6). This effect can be related to the lower hydrogen-bonding propensity of the aldehyde tails, which makes them more mobile, and to the larger free volume available to the acyl chains when the sn-2 chain is shorter.

Display large version of this figure

Figure 7

(A) Deuterium order parameter in the sn-2 lipid chains of PLPC and each oxidized lipid, in the simulations with 11% oxidized lipids. (B) Deuterium order parameter of the sn-1 chain of PLPC in the presence of 9-tc and (C) 9-al at different concentrations of oxidized lipid. Error bars were omitted in panels B and C, for the sake of clarity.

Table 6 Average deuterium order parameter of the sn-1 and sn-2 chain of PLPC and the oxidized lipids in all the simulations

		Concentration of oxidized lipids (%)	Avg SCDsn-1 PLPC	Avg SCDsn-2 PLPC	Avg SCDsn-1 ox	Avg SCDsn-2 ox
	Pure PLPC		0.168	0.129
		2.8	0.161	0.126	0.176	0.023
		5.6	0.163	0.129	0.164	0.020
	9-tc	11.1	0.159	0.126	0.161	0.023
		25.0	0.156	0.119	0.152	0.030
		50.0	0.149	0.111	0.151	0.035
		2.8	0.160	0.123	0.154	−0.016
		5.6	0.162	0.126	0.164	0.017
	13-tc	11.1	0.161	0.117	0.168	−0.011
		25.0	0.157	0.117	0.161	0.006
		50.0	0.145	0.107	0.116	0.032
		2.8	0.152	0.123	0.156	−0.033
		5.6	0.157	0.118	0.169	−0.034
	9-al	11.1	0.145	0.114	0.149	−0.030
		25.0	0.124	0.101	0.139	−0.027
		50.0	0.106	0.082	0.107	−0.013
		2.8	0.166	0.120	0.173	0.015
		5.6	0.161	0.120	0.139	0.015
	12-al	11.1	0.146	0.119	0.150	0.010
		25.0	0.135	0.109	0.136	0.026
		50.0	0.132	0.091	0.124	0.037

Wratten et al. 18 measured the membrane ordering in PLPC and DLPC bilayers containing hydroperoxide and hydroxide groups using angle-resolved fluorescence depolarization. Their results showed that the presence of oxidized lipid molecules cause a decrease in membrane order. However, other studies showed an increase 30,34,35,36 or no change 37 in the order parameter. It has been suggested that this discrepancy depends on the presence of numerous oxidative products, different from our case.

Together with the changes in the structural properties of the bilayer, dynamic properties of the lipids were also modified. We calculated the lateral diffusion coefficient from the mean-square displacements (MSD) of the lipids as a function of time. We observed that the two monolayers move relative each other and relative to water during the simulations. Both types of motion are artifacts due to the finite size of the simulated systems and to periodic boundary conditions 80. We therefore subtracted the center of mass motion of each monolayer before calculating the MSD. We then utilized the model proposed by Wohlert and Edholm for the calculation of the lateral diffusion coefficient at short and long timescales 80. This model considers two different types of diffusion occurring on different timescales. The diffusion at short times (described by the D₁ coefficient) takes place within a circular area of radius R and is not Brownian, while the diffusion at long times (described by the D₂ coefficient) involves large displacements and is Brownian. Both diffusion coefficients can be calculated by fitting the MSD curve to the following expression 80:

(4)

To estimate the error of diffusion coefficients, we split our 100-ns trajectories in two intervals of 50-ns each and fitted the four MSD curves (two curves for each leaflet) between 0 and 10ns. For the pure PLPC bilayer we found values of (10.4±0.4)×10⁻⁷cm²/s for D₁, 0.27nm² for and (0.52±0.03)×10⁻⁷cm²/s for D₂. Previous simulations of a PLPC bilayer gave a diffusion coefficient of (1.3±0.3)×10⁻⁷cm²/s for the diffusion at long times 56. This value is significantly higher than our result, but it was obtained at a higher temperature (310K). Our results compare favorably to previous simulation results for a DMPC bilayer, with D₁=13×10⁻⁷cm²/s, D₂=0.79×10⁻⁷cm²/s, and =0.3nm²80. The agreement with experimental diffusion coefficients is also very reasonable: the diffusion coefficient at short times can be compared with results from neutron scattering experiments, measuring D=(1∼10)×10⁻⁷cm²/s 81,82, while D₂ can be compared with diffusion coefficients from fluorescence recovery after photobleaching experiments, which typically give values around (0.5∼1)×10⁻⁷cm²/s 83.

Short-time and long-time diffusion coefficients for all the simulated systems are reported in Table 7. Both the short time and the long time diffusion coefficients do not change significantly with increasing concentration of oxidized lipids. Early experimental studies suggested an increase of membrane microviscosity upon peroxidation 30,31,32,84, while more recent results indicated that the effect of lipid oxidation causes pronounced structural effects but minimal effects on the membrane dynamics 18,29. Our study suggests that, while the presence of oxidized lipids has a large influence on structural properties, its effect on lipid lateral diffusion is relatively small. Long time diffusion coefficients of PLPC and oxidized lipids as a function of the concentration of oxidized lipids are reported in Fig. 8. Diffusion appears to be faster for aldehyde lipids compared to hydroperoxide lipids. This is consistent with the lower degree of ordering observed in the presence of aldehyde lipids (see Figure 7BC), and can be related to the stronger hydrogen-bonding interactions observed for hydroperoxide lipids (involving not only water but also polar headgroups of neighboring lipids) and the larger free volume available for the acyl chains in the presence of aldehyde lipids.

Table 7 Short time diffusion coefficients, long time diffusion coefficients and r0 for PLPC and for the oxidized lipids as a function of the concentration of oxidized lipids; r0 is defined as in Eq. (4)

			D1×107(cm2/s) PLPC		D1×107(cm2/s) oxidized lipids		D2×107(cm2/s) PLPC		D2×107(cm2/s) oxidized lipids		r0 (nm) PLPC		r0 (nm) oxidized lipids
		% ox	Avg	e.e.*	Avg	e.e.*	Avg	e.e.*	Avg	e.e.*	Avg	e.e.*	Avg	e.e.*
	PLPC	0.0	10.4	0.4			0.52	0.03			0.52	0.01
		2.8	10.9	0.7	6.0	1.2	0.65	0.10	0.49	0.16	0.51	0.02	0.51	0.02
		5.6	10.3	0.3	10.3	1.7	0.44	0.04	0.42	0.09	0.51	0.01	0.42	0.01
	9-tc	11.1	10.2	0.2	7.9	0.5	0.48	0.04	0.42	0.02	0.54	0.00	0.44	0.01
		25.0	9.7	0.3	7.6	1.1	0.49	0.05	0.32	0.06	0.53	0.01	0.45	0.01
		50.0	9.8	0.4	7.5	0.6	0.42	0.07	0.24	0.04	0.52	0.01	0.44	0.01
		2.8	10.6	0.3	7.1	1.5	0.62	0.03	0.36	0.15	0.51	0.01	0.47	0.04
		5.6	10.6	0.6	6.5	0.7	0.51	0.05	0.37	0.06	0.51	0.01	0.42	0.00
	13-tc	11.1	10.3	0.5	7.6	1.1	0.53	0.08	0.34	0.06	0.54	0.02	0.41	0.02
		25.0	10.3	0.7	7.5	0.4	0.54	0.08	0.39	0.05	0.52	0.02	0.41	0.01
		50.0	8.9	0.4	6.2	0.6	0.56	0.05	0.37	0.08	0.48	0.01	0.41	0.01
		2.8	10.4	0.7	7.2	0.5	0.64	0.03	0.57	0.11	0.53	0.02	0.49	0.01
		5.6	10.2	0.3	10.7	1.6	0.66	0.04	0.57	0.09	0.53	0.01	0.44	0.01
	9-al	11.1	9.7	0.5	6.9	0.3	0.56	0.08	0.49	0.07	0.55	0.02	0.52	0.01
		25.0	9.2	0.2	8.2	0.8	0.61	0.08	0.68	0.10	0.56	0.01	0.48	0.01
		50.0	8.2	0.1	6.8	0.1	0.63	0.05	0.64	0.09	0.61	0.02	0.54	0.01
		2.8	10.8	0.7	10.7	2.2	0.61	0.05	0.49	0.06	0.52	0.02	0.44	0.02
		5.6	9.9	0.2	6.9	0.8	0.48	0.06	0.35	0.06	0.53	0.01	0.52	0.03
	12-al	11.1	10.6	0.3	8.6	0.8	0.58	0.06	0.55	0.06	0.54	0.01	0.47	0.01
		25.0	10.3	0.3	9.3	0.9	0.69	0.04	0.73	0.08	0.54	0.01	0.45	0.01
		50.0	9.7	0.5	7.6	0.3	0.76	0.08	0.77	0.08	0.55	0.02	0.49	0.00

* Error estimate.

Display large version of this figure

Figure 8

Average long time (D₂) lateral diffusion coefficients of PLPC and oxidized lipids as a function of the concentration of oxidized lipids.

Water permeation through PLPC bilayers

The presence of oxidized lipids has a profound influence on the permeability of water through PLPC bilayers. Water pores are observed in all the simulations containing 5% or more oxidized lipids, and they are relatively stable at higher oxidation levels. Fig. 9 shows a water defect in a bilayer containing 13-tc lipids at 50% concentration. Based on visual inspection of the trajectories, water defects can form independently in both leaflets and they are larger in the presence of 12-al and 13-tc. This difference correlates well with the higher area per lipid in bilayers containing 12-al and 13-tc lipids, which have polar oxygen atoms closer to t