Alteration of lipid membrane structure and dynamics by diacylglycerols with unsaturated chains

Biochimica et Biophysica Acta 1858 (2016) 253–263

Contents lists available at ScienceDirect

Biochimica et Biophysica Acta journal homepage: www.elsevier.com/locate/bbamem

Alteration of lipid membrane structure and dynamics by diacylglycerols with unsaturated chains Mohammad Alwarawrah a, Fazle Hussain b, Juyang Huang a,⁎ a b

Department of Physics, Texas Tech University, Lubbock, TX 79409, United States Department of Mechanical Engineering, Texas Tech University, Lubbock, TX 79409, United States

a r t i c l e

i n f o

Article history: Received 25 September 2015 Received in revised form 5 November 2015 Accepted 18 November 2015 Available online 1 December 2015 Keywords: Umbrella model Membrane diffusion Free energy Order parameter Headgroup spacing Condensing effect

a b s t r a c t Diacylglycerols (DAGs) with unsaturated acyl chains play many important roles in biomembranes, such as a second messenger and activator for protein kinase C. In this study, three DAGs of distinctly different chain unsaturations (i.e. di16:0DAG (DPG), 16:0–18:1DAG (POG), and di18:1DAG (DOG)) are studied using atomistic MD simulation to compare their roles in the structure and dynamics of 16:0–18:1phosphatidylcholine (POPC) membranes. All three DAGs are able to produce the so-called ‘condensing effect’ in POPC membranes: decreasing area-per-lipid, and increasing acyl chain order and bilayer thickness. Our visual and quantitative analyses clearly show that DAG with unsaturated chains induce larger spacing between POPC headgroups, compared with DAG with saturated chains; this particular effect has long been hypothesized to be crucial for activating enzymes and receptors in cell membranes. DAGs with unsaturated chains are also located closer to the bilayer/aqueous interface than DPG and are more effective in slowing down lateral diffusion of molecules. We show that DAG molecules seek the “umbrella coverage” from neighboring phospholipid headgroups — similar to cholesterol. Unlike cholesterol, DAGs also hide their chains from water by laterally inserting their chains into the surrounding. Thus, acyl chains of DAG are more spread and disordered than those of PC due to the insertion. By calculating the potential of mean force (PMF) for POPC in POPC/DAG bilayers, we found that all three DAGs can significantly increase the free energy barrier for POPC to flip-flop, but only DAGs with unsaturated chains can additionally increase the free energy of POPC desorption. © 2015 Elsevier B.V. All rights reserved.

1. Introduction Diacylglycerols (DAGs) play vital roles in lipid biosynthesis and in activating protein kinase C (PKC) [1–3]. DAG has a simple lipid structure that consists of a glycerol molecule linked to two fatty acids through ester bonds [4,5]. The fatty acids can be either saturated or unsaturated [6], as shown in Fig. 1. DAGs with one saturated chain and one unsaturated chain, such as POG or 18:0–26:4DAG, are particularly biologically relevant, since this is the main type involved in cell signaling [1,3]. DAGs affect the biophysical properties of lipid membranes primarily through their molecular shape: a small polar headgroup and a much larger nonpolar chain section. In cell membranes, the overall concentration of DAG is usually low, ranging from 1 to 8 mol% [7–9]. However, DAGs can be produced and accumulated in lipid domains [10]; thus the local concentration of DAG inside a membrane domain could be significantly higher [11–13]. DAGs can alter the intrinsic curvature of a membrane and facilitate phospholipase-C induced vesicle fusion [14]. At high concentrations, DAG can induce non-lamellar phases (i.e. HII hexagonal or

⁎ Corresponding author. E-mail address: [email protected] (J. Huang).

http://dx.doi.org/10.1016/j.bbamem.2015.11.014 0005-2736/© 2015 Elsevier B.V. All rights reserved.

cubic phase) in membranes [15–17]. Szule et al. found that the chain lengths and degree of unsaturation of both DAG and phospholipids can strongly influence the lamellar (Lα) to reverse hexagonal (HII) phase transition [18]. It is well known that cholesterol can increase the chain order of a lamellar-phase lipid bilayer. A number of experiments showed that DAG can also produce a similar effect [17,19]. López-García et al. used infrared spectroscopy to investigate the interactions among DAG, Ca2 + and di16:0phosphatidylserine (DPPS) [20]. They found that DAG and Ca2 + increase the dehydration of DPPS headgroups. Dehydration of di16:0PC (DPPC) headgroups by DAG was also observed experimentally [12]. A number of studies indicated that DAG increases the spacing between phospholipid headgroups in lipid membranes due to the interposition of DAG small headgroups [17,21, 22]. Moreover, it has been suggested that increasing the spacing between phospholipid headgroups directly contributes to the increase of PKC activity [15,21–23]. It has been found that the degree of unsaturation and the length of fatty acid chains in DAG can affect its ability to activate PKC protein [17,23]. Torrecillas et al. found that DAG with one unsaturated chain (i.e. 16:0–18:1DAG (POG)) is a better activator for PKCα than a fully saturated DAG (i.e. di16:0DAG (DPG)) [3]. This is understandable, since it has been well documented that chain unsaturation increases membrane's area-per-lipid [24] as well as the

254

M. Alwarawrah et al. / Biochimica et Biophysica Acta 1858 (2016) 253–263

a systematic comparison of DAGs of different chain unsaturations at atomistic level has not yet be performed. Furthermore, it is not clear whether DAGs with unsaturated chains also produce the condensing effect in phospholipid bilayers or whether phospholipid headgroups provide “umbrella coverage” to these DAGs. More importantly, how effective are DAGs with unsaturated chains in enlarging the spacing between phospholipid headgroup, and is the spacing directly above a DAG headgroup largest? In this study, effects of three DAGs of distinctly different chain unsaturations (i.e. DPG, POG, and DOG, see Fig. 1) on the structure and dynamics of POPC bilayers were investigated using MD simulation. We are particularly interested in the differences in their abilities to enlarge spacing between PC headgroups and have developed several analyses to visualize and quantify the differences. Also, we performed free energy calculations for binary mixtures of POPC and DAG. Our results are in good agreement with experiments: we found that the unsaturated POG and DOG are located closer to the bilayer/aqueous interface, and are more effective than DPG in enlarging spacing between PC headgroups and in reducing lateral diffusion of molecules; also adding DAG to a PC bilayer causes dehydration of PC headgroups. Our data clearly indicate that an important interaction between DAGs and phospholipids is that a DAG molecule seeks the “umbrella coverage” from neighboring phospholipid headgroups to avoid the unfavorable exposure to water, similar to cholesterol in lipid membranes. Therefore, certain small-headgroup membrane molecules (such as cholesterol, ceramides and DAGs) should compete for the umbrella coverage from large-headgroup lipids (such as PC or SM) in a lipid membrane; this explains why these small-headgroup molecules often displace each other in membrane domains [32–35]. 2. Simulation methods Fig. 1. Chemical structure of POPC (a), DPG (di16:0DAG) (b), POG (16:0–18:1DAG) (c), and DOG (di18:1DAG) (d). Atom indices, such as O2, C4, N4, and P8, are the names in the molecule structure PDB files. Red arrows define orientational vectors of DAGs.

spacing between phospholipid headgroups, which may facilitate insertion of the enzyme into the membrane [21]. Molecular dynamics (MD) simulation has increasingly become an important tool for studying lipid membranes. Recently, MD simulation was used to study the interactions between DAGs and phospholipids in lipid bilayers [25–28]. Alwarawrah et al. investigated the interactions among DPG, 16:0:18:1PC (POPC), and cholesterol [25]. They showed that DPG, like cholesterol, also produces the so-called “condensing effect” in POPC bilayers: reducing area-per-lipid and lateral diffusion of molecules, and increasing acyl chain order and bilayer thickness. However, cholesterol is more effective than DPG in producing the condensing effect, because of the former's rigid sterol ring structure. The umbrella model, originally proposed to explain the cholesterol–lipid interaction [29–31], can also explain important DPG–lipid interactions. In another MD study, Vamparys et al. developed a new surface analysis method to examine di18:1PC (DOPC)/di18:1DAG (DOG) bilayers [28]. Their analysis showed that DOG produces solvent and geometrical defects in the bilayer, and these defects increase with the number of unsaturated acyl chains in PC and with the addition of DOG. More recently, coarse-grained simulations have been used to investigate the flip-flop motions of DAG and ceramide in phospholipid bilayers [27]. The simulations showed that flip-flop motions of DAG and ceramide in the di20:4PC (DAPC) membrane are slower than that of cholesterol. The flip-flop rate increases significantly as the number of double bonds of the acyl chains increases. Bennett and Tieleman investigate the interactions of cholesterol, ceramide and DPG in an ordered raft-like bilayer (i.e. POPC/PSM/cholesterol) and in pure POPC bilayer [26]. The calculated potential of mean force (PMF) showed that cholesterol has a large free energy of exchange between POPC and the raft-like bilayers, while ceramide and DAG have modest free energies of exchange. Although experiments have showed that DAGs with unsaturated chains are better activators for PKCα than DAGs with saturated chains,

2.1. Large systems Molecular dynamics simulations of POPC/18.75% POG and POPC/ 18.75% DOG bilayers were performed at 310 K with a time step of 2.0 fs and run for 500 ns. According to the phase diagram established by Jimenez-Monreal et al. using DSC, small-angle X-ray diffraction and 31 P-NMR, POPC/POG lipid bilayers are in the lamellar phase at 310 K as long as the concentration of POG is below 40 mol% [36]. The total number of lipids in each system was kept at 512, and the water-tolipid ratio was 28.625. POPC force field was from Berger et al. [37]. POG and DOG force fields were constructed from the bonded and nonbonded parameters of GROMOS87 force field [38] implemented in GROMACS as ffgmx. The partial charges of POG and DOG were set identical to that of DPG [25]. The system was run using GROMACS 4.0.7 software package [39] in the NPT ensemble using a Nosé–Hoover [40] thermostat and Parrinello–Rahman [41] barostat methods with coupling time constants of 0.1 ps and 1.0 ps, respectively. The pressures normal and parallel to the bilayer were coupled separately at 1 bar. The LINCS algorithm was used to keep the lengths of all bonds constant. Long-range electrostatic interactions were handled with the particle mesh Ewald (PME) method [42]. The cutoff distances for Lennard–Jones interactions and electrostatic interactions were both set at 1.0 nm. All other simulation conditions were identical to that in our previous study [25]. The initial structures of POPC, POG, and DOG were constructed using the Dundee PRODRG2 Server [43]. A single POPC was placed inside a solvent box, and a short simulation was performed to relax the lipid. Afterward, a bilayer of 32 POPCs was constructed from this relaxed POPC. The 32-POPC bilayer was then used to construct POPC/18.75% POG and POPC/18.75% DOG bilayer subunits. Three randomly picked POPC molecules in each leaflet were replaced by POG or DOG to create POPC/18.75% POG and POPC/18.75% DOG bilayer subunits. The VMD program (http://www.ks.uiuc.edu/Research/vmd/) was then used to remove any bad links or overlaps that could happen after the DAG insertion [44]. Following this, the GROMACS [45] utility “genconf” was

M. Alwarawrah et al. / Biochimica et Biophysica Acta 1858 (2016) 253–263

employed to replicate systems of 32 lipids to generate 128-lipid subunits. After energy minimization and equilibration, “gencof” was used again to generate the 512-lipid POPC/18.75% POG and POPC/18.75% DOG initial structures. All initial structures went through energy minimization using both the steepest descent and the conjugate gradient algorithms; each ran for 1000 steps to eliminate any bad contacts.

255

time for each window was 200 ns, and the last 20 ns of trajectories was used for analysis. The weighted histogram analysis method (WHAM) was used to calculate the potential of mean force [50]. The errors were estimated using Bayesian bootstrapping of complete histograms with 500 bootstraps [50]. 3. Results and discussion

2.2. Small systems Each of the small systems consists of 64 lipids and 4000 water molecules. For a POPC/DAG binary mixture system (i.e. POPC/18.75% DPG, POPC/18.75% POG, or POPC/18.75% DOG), 12 POPC lipids were replaced by DPG, POG or DOG molecules. All simulations were performed with the GROMACS 4.5.5 software package [39]. All systems were run in the NPT ensemble using a V-rescale thermostat [46] and Berendsen barostat [47] methods with the coupling time constants of 0.1 ps and 1.0 ps, respectively. All other simulation conditions were identical to that for the large systems. 2.3. The umbrella sampling method (not to be confused with the umbrella model) This method was used to calculate the potential of mean force (PMF) acting on a phosphate group of POPC as a function of the distance from the center of the bilayer. PMF calculations are computationally expensive: Each system has 49 simulation windows that gradually translate a pair of POPC, initially located at the opposite leaflets of the bilayer from their equilibrium locations toward the aqueous phase, at a step size of 0.1 nm between adjacent windows. The initial structure of each window was created by pulling the pair of POPC from its equilibrium location in the bilayer to the predetermined location using the umbrella potentials with a force constant of 500 kJ mol−1 nm−2 at a pulling rate of 0.01 Å/ps [48,49]. Afterward, the pair of POPC molecules was restrained; one POPC molecule was staggered partially or completely within the bulk water and the other within the bilayer. The distances between the phosphate groups of two staggered POPC molecules and the center of the bilayer were restrained using a harmonic potential with a force constant of 3000 kJ mol−1 nm−2 [48]. The simulation run

Fig. 2 shows the snapshots of POPC/18.75% POG and POPC/18.75% DOG systems at 0, 100, 300, and 500 ns. Two initial structures (Fig. 2a and e) are composed of 4 identical subunits. Due to lateral and rotational diffusions, the positions and orientations of POG and DOG molecules changed significantly as time progresses. We found that not many DAG headgroups are in contact, since DAG–POPC contacts facilitate the “umbrella coverage” and thus are more favorable. 3.1. Area-per-lipid, bilayer thickness, and volume per lipid Area-per-lipid (Apl = 2Abox/Nlipid), volume per lipid (Vpl = (Vbox − Nwatervwater)/Nlipid), and bilayer thickness (h = Vpl/Apl) were calculated by averaging over the last 100 ns of the trajectories. The results for POPC/18.75% POG and POPC/18.75% DOG bilayers, together with the results for pure POPC and POPC/18.75% DPG bilayers from a previous study [25], are listed in Table 1. The data show that all three diacylglycerols significantly decrease the area-per-lipid and volume per lipid, and increase the bilayer thickness. The thickness of POPC/18.75% DOG is slightly larger than others, because the acyl chains of DOG are longer than that of DPG. Previously, we found that DPG produces the condensing effect in POPC bilayers [25], similar to cholesterol [51]. Based on the data in Table 1, together with the increase of acyl chain order parameter by DAG in the next section, we can conclude that all three diacylglycerols produce the condensing effect in POPC bilayers. 3.2. Acyl chain order parameter The order parameter for a CH2 group in an acyl chain is defined as SCD = (3 b cos2 θ N − 1)/2, where θ is the time-dependent angle between a C\\H bond and the bilayer normal, and bN denotes time

Fig. 2. (a to d) Snapshots of the POPC/18.75% POG system at various time points. Only the lipids on the top leaflet are shown. POPCs are shown as thin lines, and POGs (in cyan) are represented by the space-filling model. The red and white spheres are oxygen and hydrogen atoms, respectively. (e to h) Snapshots of the POPC/18.75% DOG system. DOGs (in green) are represented by the space-filling model.

256

M. Alwarawrah et al. / Biochimica et Biophysica Acta 1858 (2016) 253–263

Table 1 Area-per-lipid (Apl), bilayer thickness (h), and volume per lipid (Vpl) in various systems. Systems

Apl (nm2)

h (nm)

Vpl (nm3)

POPC/18.75% POG POPC/18.75% DOG POPC/18.75% DPGa Pure POPCa

0.540 ± 0.005 0.539 ± 0.003 0.538 ± 0.005 0.653 ± 0.005

4.39 ± 0.04 4.43 ± 0.03 4.38 ± 0.04 3.87 ± 0.03

1.185 ± 0.004 1.193 ± 0.004 1.177 ± 0.004 1.264 ± 0.004

a

Data from a previous study [25].

average of an ensemble. The theoretical possible values of order parameter range from − 0.5 to + 1.0. Fig. 3 shows the order parameters of POPC and DAG acyl chains at each carbon atom position. The three DAGs increase the order parameter of POPC acyl chains roughly by the same amount (Fig. 3a & b), correlated with nearly the same amount of reductions in area-per-lipid (Table 1). The cis double bond in sn-2 chain of POPC produces a sharp local drop in the order parameter (Fig. 3b) as a double bond makes a chain more flexible. The condensing effect of DAG results in tight packing of chains, the packing being tightest in the middle section. Thus, the order parameters have maxima in the middle section of the chain, except when there is a double bond in the chain. Experimentally, it has been well established that adding DAG to a PC bilayer increases the order parameters of PC chains. In a 2H and 31 P NMR study, Goldberg et al. showed that the addition of DAG increases the order of PC acyl chains in the region adjacent to the headgroups [17]. Schorn and March found that the order parameters of di14:0PC chains increase with the addition of di14:0DAG in the lamellar phase region [19]. Our simulation result on chain order parameters is consistent with the experimental data. Interestingly, our simulation shows that the order parameters of DAG chains are lower than those of POPC chains in all three binary mixtures, which indicates that POPC chains are more ordered than DAG chains (Fig. 3c & d). This is particularly interesting for POPC/18.75% POG system, since POPC and POG have identical chains. For example, the order parameter of POPC sn-1 chain at C6 position is about 0.29 in POPC/18.75% POG system (Fig. 3a), while the order parameter of POG sn-1 chain at the same position is only about 0.19 (Fig. 3c). Our result is in excellent agreement with a previous experimental measurement of order parameters using ESR spectroscopy. Schorn and Marsh found that the order parameters

of di14:0DAG (DMG) are consistently lower than those of di14:0PC (DMPC) in DMPC/DMG lipid bilayers [19], despite the fact that the DMG and DMPC have identical chains. For example, the order parameter of DMPC sn-1 chain at C6 position in DMPC/30% DMG mixture at 60 °C was determined to be 0.37, while the order parameter of DMG sn-1 chain at the same position was found to be 0.27 [19]. The difference between the two order parameters in our simulation is very similar to the measured difference in their system. Our data shows that such difference also exists in systems containing DAG with unsaturated chains. In a previous study, we found that DAG molecules have a tendency to laterally insert its acyl chains into surrounding chain matrix in order to reduce its effective hydrophobic cross-section [25]. This lateral extension explains why acyl chains of DAGs are more disordered than the more compact acyl chains of POPC. The average distance between acyl chains calculated in the next section also supports this explanation. As expected, the cis double bonds in both chains of DOG and in the sn-2 chain of POG all produce sharp local drops in the order parameter. Note that the order parameters in Fig. 3b–d oscillate slightly as functions of carbon atom position, this so-called odd–even effect having been observed in a number of other studies [25,52,53]. This oscillation is likely related to the alternating C\\C bond orientation in each chain. It should be pointed out that order parameters depend not only on the orders of chains, but also on the orientations of the chains. For example, a perfectly ordered acyl chain in all-trans conformation with its mean axis normal to the bilayer (i.e., θ = 90°) would result in SCD = − 0.5, while an identical chain with its mean axis parallel to the bilayer (i.e., θ = 0°) would have SCD = + 1.0 [54]. The C\\C bonds in an acyl chain typically change orientation as a function of carbon position in an alternating fashion, which would induce an oscillation in the order parameter. If a chain is highly disordered or spins rapidly, it can smooth out the odd–even effect. On the other hand, for highly ordered chains, the odd–even effect becomes noticeable. Since both DAG and cholesterol are small-headgroup molecules and both can produce the condensing effect in a lipid membrane, they should be able to complement each other in altering membrane properties and functions. The complementary actions of DAG and cholesterol have been observed in many experiments. For example, it has been shown that less DAG is needed to drive a lipid membrane from the bilayer phase to a non-lamellar phase, if cholesterol is present in the

Fig. 3. (a and b) Order parameters of POPC chains. (c and d) Order parameters of diacylglycerol chains. The standard deviations are less than 1% of the values.

M. Alwarawrah et al. / Biochimica et Biophysica Acta 1858 (2016) 253–263 Table 2 Average distances between two acyl chains of POPC, DPG, POG, and DOG in binary mixtures of POPC and diacylglycerol. Systems

POPC (nm)

DAG (nm)

POPC/18.75% POG POPC/18.75% DOG POPC/18.75% DPG Pure POPC

0.99 ± 0.02 0.98 ± 0.02 0.99 ± 0.02 1.12 ± 0.02

1.07 ± 0.03 1.08 ± 0.03 1.06 ± 0.03 –

membrane [55]. More interestingly, Armstrong and Zidovetzki found that the activation of protein kinase C by DAG is amplified by the presence of cholesterol [56]. 3.3. Distance between acyl chains In many simulation snapshots, we observed that acyl chains of some DAG molecules spread widely. Previously, we reasoned that DPG molecules can practically reduce the size of their nonpolar sections by inserting their chains into the surrounding chain matrix [25], as illustrated in Fig. 7a. We calculated the average distance between two acyl chains of a POPC and compared it with that of a DAG. The distance between C44 carbon of sn-1 chain and C24 carbon of sn-2 chain was used to represent the distance between two POPC chains (see Fig. 1); the distance between C33 carbon of sn-1 chain and C15 carbon of sn-2 chain was used to represent the distance between two DPG chains; and the distance between C35 carbon of sn-1 chain and C15 carbon of sn-2 chain was used to represent the distances between two POG or DOG chains. Table 2 shows that the addition of DPG, POG, or DOG decreases the average distance between two acyl chains of a POPC by ~10%. On the other hand, in all three systems containing DAG, the average distance between the acyl chains of DAG is larger than that between POPC chains. This shows that DAGs with unsaturated chains also tend to insert its chains into the surrounding chain matrix, similar to DPG with saturated chains. The difference in average distances explains why acyl chains of DAG have lower order parameters than acyl chains of POPC. 3.4. Electron density profiles Fig. 4 shows the electron density profiles of POPC, DPG, POG, DOG and water across the lipid bilayers in the four systems. The profiles of POPC in Fig. 4a show that the addition of any type of DAG produces a similar effect to the bilayer: The POPC profiles expand in the direction of bilayer normal, indicating an increase of bilayer thickness due to the condensing effect. Furthermore, the two peaks for POPC headgroups become sharper, indicating that POPC headgroups become more ordered and more parallel to the bilayer–aqueous interfaces [51]. The electron density profiles for DPG, POG and DOG in Fig. 4b show that at

257

the bilayer center, electron density is highest in POPC/18.75% DOG system and lowest in POPC/18.75% DPG system, correlated with the lengths of acyl chains of DAG. Fig. 4c shows that the water–bilayer interfaces for systems containing DAGs are located farther away from the bilayer center, due to the increase of bilayer thickness. Previously, infrared spectroscopy studies showed that DAGs induce dehydration of phospholipid headgroups in lipid bilayer [12,20]. To confirm this result, we calculated the average number of water molecules around a phosphate group of POPC. Water molecules located between zavg(P8POPC) + 0.5 nm and zavg(P8POPC) − 0.5 nm on each leaflet were counted and averaged over the last 100 ns of the trajectory. In the pure POPC system, the average number of water molecules per lipid is 10 ± 0.2. However, adding diacylglycerol (i.e. DPG, POG or DOG) to a POPC bilayer decreases the number to 8 ± 0.2. Thus, our simulations are consistent with experiments. Moreover, we calculated the relative positions of POPC, DPG, POG, and DOG in the bilayers. In the POPC/18.75% DPG system, on average, the C4 atom of DPG is located 1.9 Å below the C13 atom of POPC [25]. On the other hand, the C4 atoms of POG and DOG in POPC bilayers are located ~ 1.6 Å and ~1.4 Å below the C13 atom of POPC, respectively. Thus, DAGs with unsaturated chains, POG and DOG, are located closer to bilayer/aqueous interface than the fully saturated DPG.

3.5. The increase of spacing between PC headgroups Previously, we showed that DPG indeed increases the average distance between POPC headgroups [25]. The largest spacing occurs between the first and the second nearest-neighbor PC headgroups from a DAG (Fig. 7b). In order to directly visualize the spacing increase, we constructed two-dimensional (2D) number density maps of molecules, which include atoms from both POPC and DAG, in the upperleaflet headgroup regions for all four systems (Fig. 5). The peak position of POPC electron density profile in the upper-leaflet (i.e., the left peak in Fig. 4a) was taken as the center position of POPC headgroup for each system. The number density map was calculated to include all atoms located within a 2 nm thick slab centered at the peak. In the POPC bilayer without DAG, the 2D number density in the headgroup region is relatively high (Fig. 5a). In the presence of DAGs, because DAG's headgroup is far smaller than the headgroup of POPC, the number density decreases in the headgroup region (Fig. 5b, c, and d). Particularly, in POPC/18.75% POG and POPC/18.75% DOG systems (Fig. 5c and d), there are some locations with very low number density (i.e., with dark blue color), corresponding to large gaps between POPC headgroups. Therefore, Fig. 5 clearly illustrates that DAGs with unsaturated chains, POG and DOG, are more effective in creating large spacing between POPC headgroups, because of their larger hydrophobic cross-sectional area (i.e., the cross-sectional area of DAG in the acyl-chain region).

Fig. 4. The electron density profiles of POPC (a), DPG, POG, and DOG (b), and water (c) across the bilayer.

258

M. Alwarawrah et al. / Biochimica et Biophysica Acta 1858 (2016) 253–263

Fig. 5. 2D number density maps for the headgroup region show DAGs creating gaps between POPC headgroups. (a) POPC; (b) POPC/18.75% DPG; (c) POPC/18.75% POG; and (d) POPC/ 18.75% DOG.

To gain more insights, we used 2D RDFs to investigate the distributions of POPC headgroups and acyl chains around DAGs. For the pure POPC system, two RDFs were calculated, one between POPC headgroups and the other between POPC headgroup and the center of mass (COM) of POPC chains (Fig. 6a). The midpoint of its P–N vector was simply taken as the POPC headgroup position, and the COM of two acyl chains from the same POPC above the double bond location was used as the position of POPC chains. The RDF between POPC headgroups, the black solid line in Fig. 6c, shows that very few neighboring POPC headgroups are within 0.49 nm from a POPC headgroup. This is expected since a POPC does not need the umbrella coverage from neighboring POPC headgroups. For each system containing DAGs, one RDF was calculated between DAG headgroup and POPC headgroup, and the other was calculated between DAG headgroup and COMs of POPC chains (see

Fig. 6b). In Fig. 6c, the RDF distributions of POPC headgroups (i.e., P–N vectors) around diacylglycerol O2 oxygen (i.e. DPG, POG, and DOG headgroups) have large peaks at ~ 0.29 nm and small peaks at ~ 0.48 nm. These two peaks presumably correspond to the nearestneighbor POPC headgroups extending toward DAG headgroups to provide the umbrella coverage. Two close peaks indicate that certain orientations and locations of the POPC headgroup maybe preferred. In the same distance range, the RDFs of POPC acyl chains are significantly lower than the RDFs of POPC headgroups, clearly indicating that the POPC headgroups next to a DAG extend toward the latter's headgroup. At the distances approximately from 0.4 nm to 1.0 nm, the RDFs of POPC chains are higher than the RDFs of POPC headgroups, showing that there are more POPC chains than POPC headgroups in this distance interval. Thus, the spacing between PC headgroups in this region is

Fig. 6. 2D RDFs illustrating the distributions of headgroups and acyl chains of POPC around POPC and DAGs in pure POPC, POPC/18.75% DPG, POPC/18.75% POG, and POPC/18.75% DOG bilayers. The position of a POPC headgroup was taken as the midpoint of its P–N vector, and the center of mass of two acyl chains from the same POPC above the double bond was used as the position of POPC chains.

M. Alwarawrah et al. / Biochimica et Biophysica Acta 1858 (2016) 253–263

larger, compared to that in a pure POPC bilayer. For distances larger than 1.2 nm, the difference between two RDFs gradually vanishes and both RDFs approach the normal value of 1. Furthermore, we introduced the numerical parameter, “Spacing Index” (SI), to quantify the strength of the spacing opening between PC headgroups caused by DAGs. For example, SI for DPG is defined as SI ≡ ∫ [RDF(O2DPG − CHAINSPOPC) − RDF(O2DPG − PNPOPC)] dx for x approximately from ~ 0.4 to ~ 1.0 nm, at the two distances where two RDFs cross. Visually, SI is the area between the two corresponding RDFs between the two crossing points located near 0.4 and 1.0 nm in Fig. 6c. A larger SI value would indicate a larger difference between two RDFs and consequently a larger spacing between POPC headgroups. The calculated SI for POPC/18.75% DOG system is 0.055, which is larger than the SI for POPC/18.75% POG system (0.050) and for POPC/18.75% DPG system (0.041). In comparison, the SI for pure POPC system is only 0.009 (using the two crossing points located near 0.6 and 0.8 nm). Thus, the spacing between POPC headgroups is much larger in systems containing DAG. Our result indicates that all three DAGs can increase the spacing between POPC headgroups; however, DOG and POG molecules create larger spacing between POPC headgroups than DPG molecules do. The difference is caused by the double bonds on acyl chains of DOG and POG, which increase the hydrophobic cross-sectional area of DAG. Our result agrees with previous experiments, which show that POG is more effective than DPG in activating PKCα protein [3]. We hypothesize that POG and DOG increase the spacing — between the first and the second nearest-neighbor PC headgroups — as illustrated in Fig. 7b [25], which facilitates the docking of the PKC α-domain. 3.6. The umbrella effect Whether POPC headgroups also provide “umbrella coverage” to DAGs with unsaturated chains has not been investigated previously. To quantitatively measure the umbrella effect and the preferred orientation of POPC headgroups, two 2D radial distribution functions (RDF) were calculated in the plane parallel to the bilayer surface for each POPC/diacylglycerol system: the first between the O2 atom of DAG and the N4 atom of PC's choline group, and the second between the O2 and PC's phosphate residue P8 (see white dashed lines in Fig. 7a). Our result shows that POPC headgroups also provide the umbrella coverage for POG and DOG, similar to the coverage provided to DPG and cholesterol [25,57]. As shown in Fig. 7c, the RDF of O2DAG − N4POPC is higher than the RDF of O2DAG − P8POPC for the distance less than 0.25 nm from a DAG headgroup, and the relation is reversed for the distance between 0.25 and 0.45 nm. Thus, the first layer of POPC headgroups (i.e., P–N vectors) is within ~ 0.45 nm from a DAG. Based on the area-per-lipid data in Table 1, the estimated average radius of each molecule is about 0.41 nm. Clearly, the POPC headgroups next to

259

a DAG extend over the DAG and point their P–N vectors toward the DAG (Fig. 7a). Using the same analysis, the second layer of POPC headgroups, located roughly between 0.45 and 1.2 nm from a DAG headgroup, also statistically points their P–N vectors towards the DAG. We calculated the Umbrella Index [UI ≡ ∫ | RDF(O2DAG − N4PC) − RDF(O2DAG − P8PC)| dx for x from 0 to 3 nm] for POG and DOG and compared it with the previous result for DPG [25]. The UI for POPC/18.75% DOG system is 0.269, which is larger than the UI for POPC/18.75% POG system (0.245) and for POPC/18.75% DPG system (0.239). This result indicates that the umbrella coverage is highest for DOG and lowest for DPG, directly related to the chain unsaturation of DAGs. Pandit et al. used a bivariate correlation function to show the nonuniform angular distribution of sphingomyelin around a cholesterol molecule [58]. Here we apply a similar technique to visualize the angular distribution of POPC P–N vectors around a DAG. We first calculated 2D RDF(O2DAG − N4PC)(r, φ) and RDF(O2DAG − P8PC)(r, φ) as functions of distance r and polar angle φ. For each DPG molecule, the direction of the vector pointing from C23 to O5 atom was taken as φ = 0, and for POG and DOG the direction of the vector is from C25 to O5 (see Fig. 1). Then, the difference between the two RDFs in each system (i.e., RDF(O2DAG − N4PC)(r, φ) − RDF(O2DAG − P8PC)(r, φ)) was plotted as a 2D contour on the XY-plane of DAG body coordinate system in Fig. 8. In Fig. 8, the centers of contour plots are covered by yellow-tored color (i.e., the difference is positive), indicating many N4 atoms of neighboring POPC headgroups are either directly over or right next to a DAG headgroup. A little farther away from the centers, the color turns green-to-blue (i.e., the difference is negative), indicating an elevated concentration of P8 atoms of POPC. However, the distribution of green-to-blue color (i.e., P8 atoms) is not uniform. For the POPC/ 18.75% POG and POPC/18.75% DOG systems, two blue-color short arcs are located at ~ 0 and ~180° angles, indicating two preferred locations for P8 atoms of nearest-neighbor POPC headgroups. At a slightly farther distance, there are two yellow-to-red color arcs located at ~ 75 and ~285° angles, indicating the preferred locations for the N4 atoms from 2nd-nearest-neighbor POPC headgroups. For the POPC/18.75% DOG system (Fig. 8c), there are two additional dark-blue-color short arcs located at ~ 110 and ~185° angles, indicating that the packing of POPC headgroups surrounding DOGs is more structured (i.e., less random). Our results clearly indicate that all three types of DAGs receive the umbrella coverage from surrounding POPC headgroups. Our simulations show that DAGs need the umbrella coverage from PC headgroups, similar to cholesterol. Ceramide is another membrane molecule with a small headgroup and a large hydrophobic body. Pandit et al. performed a comparative study of C16:0-ceramide and cholesterol in POPC bilayers [59]. They found that the effects of C16:0-ceramide are very similar to cholesterol: decreasing area-per-lipid and increasing the order of POPC acyl chains. However, cholesterol is more effective in producing the condensing effect [59]. A natural consequence of

Fig. 7. (a) Conceptual illustration of the lateral insertion of DAG chains (in green) and the umbrella coverage on DAG by a neighboring POPC headgroup. (b) Conceptual illustration of increasing spacing between PC headgroups (blue) induced by a DAG molecule (green). The largest space is usually between the first and the second nearest-neighbor PC headgroups from a DAG. (c) 2D RDFs between diacylglycerol oxygen O2 and nitrogen N4 in POPC choline group or phosphorus P8 in POPC phosphate group.

260

M. Alwarawrah et al. / Biochimica et Biophysica Acta 1858 (2016) 253–263

Fig. 8. 2D contour plots of RDF(O2DAG − N4PC)(r, φ) − RDF(O2DAG − P8PC)(r, φ) on the XY-plane of DAG body coordinate system in POPC/18.75% DPG (a), POPC/18.75% POG (b), and POPC/ 18.75% DOG (c) bilayer systems. The direction of a DAG orientation vector (see Fig. 1) was taken as φ = 0.

these results is that some small-headgroup membrane molecules (such as cholesterol, ceramides and DAGs) should compete for the umbrella coverage from large-headgroup lipids (such as PC or SM) in a lipid membrane. A number of experiments have shown that it is indeed the case. Megha and London found that ceramide preferentially partitions into lipid raft domains and displaces cholesterol from lipid rafts [33]. Others have shown that both DAG and ceramide can activate cholesterol by displacing cholesterol from ordered membrane domains [32,34,35]. Ali et al. found that ceramide can drive cholesterol out of ordered lipid bilayer phase into cholesterol crystal phase at high cholesterol concentrations, and the replacement ratio is precisely 1 to 1 [60]. These experimental results support our general conclusion. Cholesterol has a rigid sterol ring structure, and the primary way to shield its hydrophobic body from unfavorable interactions with water is through the umbrella coverage by surrounding lipid headgroups. The umbrella model quantitatively described the values of maximum solubility of cholesterol in lipid bilayers [31,61], explained the driving force of lipid raft formation [62], and successfully predicted the chemical potential profiles of cholesterol [29,30]. Bjorkbom et al. experimentally tested the umbrella model and found that the size of sphingomyelin (SM) headgroup is crucial for SM–cholesterol interaction, in agreement with the model [63]. However, in a recent study, Artetxe et al. found that

the size of the SM headgroup had no marked effect on the thermal stability of ordered domains formed by SM/ceramide or SM/DPG interactions [64]. They hypothesized that interfacial hydrogen bonding and van der Waals' attraction forces between SM and ceramide could contribute to the formation of ordered domains. Our simulation shows that DAG molecules, unlike cholesterol, do not solely rely on the umbrella coverage: a DAG also hides its hydrophobic acyl chains from water by laterally inserting its chains into surrounding PC chain matrix, as illustrated in Fig. 7a. The insertion increases the distance between two acyl chains of DAG (Table 2) and decreases the order parameters of DAG chains (Fig. 3). This essential difference between cholesterol and DAG could explain the above experimental observation [63,64]. 3.7. Lateral diffusion of lipids The mean square displacements (MSD or b r(t)2 N) of POPC, DAGs and water molecules as functions of time are plotted in Fig. 9. The systems were allowed to relax for the first 100 ns, and the calculations were performed for the remaining time. The curves for water are very straight because of excellent statistical sampling. All other curves have three distinct sections: an initial rapid rise, a nearly linear portion, and a noisy section at end. The initial rapid rise (100–105 ns) of MSD is

Fig. 9. The mean square distance (MSD) of POPC (a), DPG, POG, and DOG (b), and water (c) vs. time. The calculations started at 100 ns.

M. Alwarawrah et al. / Biochimica et Biophysica Acta 1858 (2016) 253–263

261

Table 3 Lateral diffusion coefficients of POPC, diacylglycerols (i.e. DPG, POG, and DOG), and water (SOL) in binary mixtures of POPC and diacylglycerol. Lateral diffusion coefficient (10−8 cm2/s) Systems

POPC

DAG

SOL

POPC/18.75% POG POPC/18.75% DOG POPC/18.75% DPGa Pure POPCa

3.23 ± 0.01 2.90 ± 0.01 8.34 ± 0.01 11.44 ± 0.01

3.34 ± 0.01 3.03 ± 0.01 8.11 ± 0.01 –

3539.64 ± 0.06 3553.16 ± 0.07 3539.59 ± 0.17 3119.45 ± 0.04

a

Data from a previous study [25].

primarily due to the local motions of molecules; the linear portion reflects the true lateral diffusion behavior; and the noisy section at end is due to poor statistics (i.e., fewer occurrences) for large displacements. Lateral diffusion coefficients (D = 1/4d b r(t)2 N/dt) were obtained by fitting the linear portion of mean square distance (MSD) curves in Fig. 9, using the least-squares method. The MSD curves were calculated for the time periods of 100–500 ns and 150–500 ns, and D was calculated using the average of the linear fits to 120–170 ns and 170–220 ns sections of the MSD curves. The calculated diffusion coefficients are listed in Table 3. Adding DPG to POPC bilayer decreases the diffusion coefficient of POPC from 11.44 × 10−8 cm2/s to 8.34 × 10−8 cm2/s [25], i.e. by 27%. However, DAGs with unsaturated chains are far more effective in slowing down the lateral diffusion of lipids. The calculated diffusion coefficients of POPC in the POPC/18.75% POG and POPC/18.75% DOG systems are 3.23 × 10− 8 and 2.90 × 10− 8 cm2/s, respectively; these are 2 to 3 times smaller than that in POPC/18.75% DPG bilayer. This is not surprising since the lateral diffusion of lipids is known to be slower in a bilayer containing unsaturated chains. For example, the lateral diffusion coefficients for DPPC, POPC, and DOPC are ~ 12–13.6 × 10− 8 cm2/s (at 321 K) [65,66], 8.6 × 10−8 cm2/s (at 322 K) [67], and 6.3 × 10−8 cm2/s (at 298 K) [68], respectively. The cis double bonds in acyl chains of POG and DOG often cause kinks in the chains. The kinks produce many complex physical contacts between DAG and POPC chains in the already tightly packed bilayers; the entanglement of DAG and POPC slows down the lateral diffusions of both DAG and POPC.

3.8. Potential mean force (PMF) The PMFs of POPC in the pure POPC, POPC/18.75% DPG, POPC/18.75% POG, and POPC/18.75% DOG bilayer systems as functions of location are shown in Fig. 10. In order to compare different systems, we choose a common state, i.e., a POPC molecule in the water phase, away from the bilayer, as our reference state. During the umbrella sampling simulations, we staggered the phosphate groups of a pair of POPC in all simulated windows [48,49]. The PMFs of POPC in all four bilayers have deep free energy minima at the equilibrium positions of POPC in the bilayer. The addition of any type of diacylglycerol (i.e. DPG, POG, or DOG) to a POPC bilayer moves the positions of free energy minima farther away from the bilayer center. The positions of minima differ slightly between systems, due to the differences in bilayer thickness. The free energy barrier for POPC flip-flop (ΔGflip-flop), which is defined as the difference in

Table 4 The free energy barriers for flip-flop (ΔGflip-flop) and the free energies for desorption (ΔGDesorption) in binary mixtures of POPC and diacylglycerol. Systems

ΔGflip-flop (kJ/mol)

ΔGDesorption (kJ/mol)

POPC/18.75% POG POPC/18.75% DOG POPC/18.75% DPG Pure POPC

115.9 ± 3.6 118.9 ± 3.6 110.6 ± 3.0 88.4 ± 4.9

86.6 ± 4.2 83.6 ± 3.8 78.3 ± 3.6 78.5 ± 5.3

Fig. 10. Potential mean forces (PMFs) of POPC in pure POPC, POPC/18.75% DPG, POPC/ 18.75% POG, and POPC/18.75% DOG bilayer systems. The PMFs were set to zero in the common reference state, in which a POPC is located in the aqueous phase, away from the bilayer. Estimated error bars were calculated using Bayesian bootstrapping of complete histograms with 500 bootstraps.

free energy at the bilayer center and at the equilibration minima, as well as the free energy for desorption (ΔGDesorption), which is defined as the difference in free energy at the aqueous phase and at the equilibrium minima, are shown in Table 4. In the pure POPC system, the free energy for desorption is 78.5 kJ/mol, which is close to that for pure DPPC at 323 K (80 kJ/mol) previously calculated by MD simulations [48,49]. Adding DPG to POPC bilayer does not seem to affect ΔGDesorption. However, adding DAG with unsaturated chains increases ΔGDesorption: The calculated ΔGDesorption for POPC/18.75% POG and POPC/18.75% DOG bilayers is 86.6 kJ/mol and 83.6 kJ/mol, respectively. This result indicates that POG is more favorable to POPC than other DAGs, possibly due to the match of acyl chains with POPC. Table 4 and Fig. 10 show that adding DAG to a POPC bilayer increases the free energy barrier for flip-flop significantly and the increase is largely due to the increase of free energy value at the center of the bilayer. This is similar to adding cholesterol to DPPC bilayer, which increases the free energy barrier [48]. DAGs increase the order, thickness, and compactness of the bilayers, similar to cholesterol. Therefore, DAGs make more difficult for a POPC to flip-flop across the bilayer. Furthermore, Table 4 shows that the more double bonds the DAG has, the higher the ΔGflip-flop, likely due to the entanglement of POPC with DAG's unsaturated chains. In the process of calculating PMF, in one of the simulation windows, we placed a POPC at the bilayer center in the pure POPC system. We found that this POPC induces a water pore across the bilayer as shown in Fig. 11a. Since we observed the pore formation, the free energy for pore formation in a pure POPC bilayer should be less or equal to ΔGflip-flop, 88.4 kJ/mol. We did not observe a pore formation when a POPC was placed at the center of the POPC/18.75% DPG, POPC/18.75% POG, or POPC/18.75% DOG bilayers (Fig. 11b, c & d). Therefore, we can presume that the free energy for pore formation would be greater than ΔGflip-flop in these systems, i.e., 110.6, 115.9, and 118.4 kJ/mol in the POPC/18.75% DPG, POPC/18.75% POG, and POPC/18.75% DOG bilayers, respectively. It would be much more difficult to form pores in cellular membranes containing DAGs, due to their condensing effect. Fig. 11b, c & d also show that a few water molecules were pulled into the bilayer solvated POPC's headgroup at the center of POPC/18.75% DPG, POPC/18.75% POG and POPC/18.75% DOG bilayers. Bennett et al. used MD simulation to calculate the free energy for pore formation in pure DPPC and DPPC/cholesterol systems [48]. The free energy of pore formation in pure DPPC was 80 kJ/mol, and for DPPC with 20% cholesterol was 106 kJ/mol. This shows that molecules with small headgroups,

262

M. Alwarawrah et al. / Biochimica et Biophysica Acta 1858 (2016) 253–263

Fig. 11. Final snapshots of the simulations started from the initial states, in which the phosphate of a POPC was placed at the center of the bilayer. (a) Pure POPC: A large water pore spanning both leaflets was observed. (b) POPC/18.75% DPG. (c) POPC/18.75% POG. (d) POPC/18.75% DOG. No water pore was observed in (b)–(d); however, seven to eight water molecules were pulled into the bilayer solvated the POPC's headgroup. POPC in cyan (P8 and N4 in gold and blue, respectively), DPG in green, POG in light blue, DOG in orange, and water in red.

such as cholesterol and DAG, increase the free energy of pore formation in cellular membranes due to the condensing effect. Note that PC flipflop is very slow in cell membranes, and the pore formation observed in our simulation was likely induced by artificially placing a POPC at the bilayer center, which was required for calculating PMF. However, the observation allowed us to estimate the value of ΔGflip-flop, and the same approach has been previously used by others [48]. Although the overall concentrations of DAG in many cell membranes are quite low, the local concentrations of DAG inside some membrane domains where DAGs are produced can be significantly higher [11– 13]. Thus, the results from this study at 18.75% of DAG should be biologically relevant. Even at a lower DAG concentration, DAG molecules would still need the “umbrella coverage” from surrounding largeheadgroup lipids, since it is driven by the molecular shape of DAGs and by the fundamental hydrophobic interaction. 4. Conclusions In this study, our MD simulations confirmed and mostly explained a number of important experimental results. Compared with DAGs with fully saturated chains, DAGs with unsaturated chains are far more effective in enlarging spacing between POPC headgroups, which can facilitate the docking of PKCα domains in biomembranes. Furthermore, DAGs with unsaturated chains are more effective in slowing down the lateral diffusion of molecules, increasing the free energy of lipid desorption, and forcing neighboring PC headgroups to provide the “umbrella coverage”. The above effects can be attributed to the fact that DAGs with unsaturated chains effectively have larger hydrophobic crosssectional area, due to the kinks in their acyl chains caused by the cis double bonds. In general, a DAG has a small polar headgroup and a relatively large nonpolar acyl chain section. The common molecular shape of DAGs and cholesterol determines that DAG molecules can also produce the so-called cholesterol condensing effects in a lipid bilayer: decreasing area-per-lipid, increasing acyl chain order and bilayer thickness, and increasing free energy barrier for lipid flip-flop. Our results demonstrate that an important interaction between DAGs and phospholipids is that a DAG molecule needs the “umbrella coverage” from neighboring phospholipid headgroups, similar to cholesterol. In addition, DAGs also have a tendency to laterally insert their chains into the surrounding, in order to reduce their effective hydrophobic cross-sectional area.

Transparency Document The Transparency document associated with this article can be found, in the online version. Acknowledgments The authors acknowledge the High Performance Computing Center (HPCC) at Texas Tech University for providing HPC resources for this work. Appendix A. Supplementary data POG.itp and DOG.itp are GROMACS force fields for 16:0–18:1DAG and di18:1DAG, respectively. Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.bbamem.2015.11.014. References [1] J.C. Gomez-Fernandez, S. Corbalan-Garcia, Diacylglycerols, multivalent membrane modulators, Chem. Phys. Lipids 148 (2007) 1–25. [2] Y.V. Shulga, M.K. Topham, R.M. Epand, Regulation and functions of diacylglycerol kinases, Chem. Rev. 111 (2011) 6186–6208. [3] A. Torrecillas, S. Corbalan-Garcia, A. de Godos, J.C. Gomez-Fernandez, Activation of protein kinase C alpha by lipid mixtures containing different proportions of diacylglycerols, Biochemistry 40 (2001) 15038–15046. [4] S. Carrasco, I. Merida, Diacylglycerol, when simplicity becomes complex, Trends Biochem. Sci. 32 (2007) 27–36. [5] F.M. Goni, A. Alonso, Structure and functional properties of diacylglycerols in membranes, Prog. Lipid Res. 38 (1999) 1–48. [6] M.J. Wakelam, Diacylglycerol—when is it an intracellular messenger? Biochim. Biophys. Acta 1436 (1998) 117–126. [7] Z.F. Chang, D.Y. Huang, Decline of protein-kinase-C activation in response to growth-stimulation during senescence of Imr-90 human-diploid fibroblasts, Biochem. Biophys. Res. Commun. 200 (1994) 16–27. [8] R.M. Kramer, G.C. Checani, D. Deykin, Stimulation of Ca-2+-activated humanplatelet phospholipase-A2 by diacylglycerol, Biochem. J. 248 (1987) 779–783. [9] A. Wolfman, I.G. Macara, Elevated levels of diacylglycerol and decreased phorbol ester sensitivity in Ras-transformed fibroblasts, Nature 325 (1987) 359–361. [10] A.K. Rouquette-Jazdanian, C. Pelassy, J.P. Breittmayer, J.L. Cousin, C. Aussel, Metabolic labelling of membrane microdomains/rafts in Jurkat cells indicates the presence of glycerophospholipids implicated in signal transduction by the CD3 T-cell receptor, Biochem. J. 363 (2002) 645–655. [11] T. Heimburg, U. Wurz, D. Marsh, Binary phase-diagram of hydrated dimyristoylglycerol– dimyristoylphosphatidylcholine mixtures, Biophys. J. 63 (1992) 1369–1378.

M. Alwarawrah et al. / Biochimica et Biophysica Acta 1858 (2016) 253–263 [12] F. Lopez-Garcia, J. Villalain, J.C. Gomez-Fernandez, P.J. Quinn, The phase behavior of mixed aqueous dispersions of dipalmitoyl derivatives of phosphatidylcholine and diacylglycerol, Biophys. J. 66 (1994) 1991–2004. [13] K.A. Riske, H.G. Dobereiner, Diacylglycerol-rich domain formation in giant stearoyloleoyl phosphatidylcholine vesicles driven by phospholipase C activity, Biophys. J. 85 (2003) 2351–2362. [14] A.V. Villar, F.M. Goni, A. Alonso, Diacylglycerol effects on phosphatidylinositolspecific phospholipase C activity and vesicle fusion, FEBS Lett. 494 (2001) 117–120. [15] S. Das, R.P. Rand, Modification by diacylglycerol of the structure and interaction of various phospholipid bilayer membranes, Biochemistry 25 (1986) 2882–2889. [16] R.M. Epand, R.F. Epand, C.R. Lancaster, Modulation of the bilayer to hexagonal phase transition of phosphatidylethanolamines by acylglycerols, Biochim. Biophys. Acta 945 (1988) 161–166. [17] E.M. Goldberg, D.S. Lester, D.B. Borchardt, R. Zidovetzki, Effects of diacylglycerols and Ca2+ on structure of phosphatidylcholine/phosphatidylserine bilayers, Biophys. J. 66 (1994) 382–393. [18] J.A. Szule, N.L. Fuller, R.P. Rand, The effects of acyl chain length and saturation of diacylglycerols and phosphatidylcholines on membrane monolayer curvature, Biophys. J. 83 (2002) 977–984. [19] K. Schorn, D. Marsh, Lipid chain dynamics and molecular location of diacylglycerol in hydrated binary mixtures with phosphatidylcholine: spin label ESR studies, Biochemistry 35 (1996) 3831–3836. [20] F. Lopez-Garcia, V. Micol, J. Villalain, J.C. Gomez-Fernandez, Infrared spectroscopic study of the interaction of diacylglycerol with phosphatidylserine in the presence of calcium, Biochim. Biophys. Acta 1169 (1993) 264–272. [21] E.J. Bolen, J.J. Sando, Effect of phospholipid unsaturation on protein kinase C activation, Biochemistry 31 (1992) 5945–5951. [22] S.J. Slater, M.B. Kelly, F.J. Taddeo, E. Rubin, C.D. Stubbs, Evidence for discrete diacylglycerol and phorbol ester activator sites on protein kinase C. Differences in effects of 1-alkanol inhibition, activation by phosphatidylethanolamine and calcium chelation, J. Biol. Chem. 269 (1994) 17160–17165. [23] E.M. Goldberg, D.S. Lester, D.B. Borchardt, R. Zidovetzki, Effects of diacylglycerols on conformation of phosphatidylcholine headgroups in phosphatidylcholine/ phosphatidylserine bilayers, Biophys. J. 69 (1995) 965–973. [24] J.F. Nagle, S. Tristram-Nagle, Structure of lipid bilayers, Biochim. Biophys. Acta Rev. Biomembr. 1469 (2000) 159–195. [25] M. Alwarawrah, J. Dai, J. Huang, Modification of lipid bilayer structure by diacylglycerol: a comparative study of diacylglycerol and cholesterol, J. Chem. Theory Comput. 8 (2012) 749–758. [26] W.F. Bennett, D.P. Tieleman, Molecular simulation of rapid translocation of cholesterol, diacylglycerol, and ceramide in model raft and nonraft membranes, J. Lipid Res. 53 (2012) 421–429. [27] F. Ogushi, R. Ishitsuka, T. Kobayashi, Y. Sugita, Rapid flip-flop motions of diacylglycerol and ceramide in phospholipid bilayers, Chem. Phys. Lett. 522 (2012) 96–102. [28] L. Vamparys, R. Gautier, S. Vanni, W.F. Bennett, D.P. Tieleman, B. Antonny, C. Etchebest, P.F. Fuchs, Conical lipids in flat bilayers induce packing defects similar to that induced by positive curvature, Biophys. J. 104 (2013) 585–593. [29] M.R. Ali, K.H. Cheng, J. Huang, Assess the nature of cholesterol–lipid interactions through the chemical potential of cholesterol in phosphatidylcholine bilayers, Proc. Natl. Acad. Sci. U. S. A. 104 (2007) 5372–5377. [30] J. Huang, Model membrane thermodynamics and lateral distribution of cholesterol: from experimental data to Monte Carlo simulation, Methods Enzymol. 455 (2009) 329–364. [31] J. Huang, G.W. Feigenson, A microscopic interaction model of maximum solubility of cholesterol in lipid bilayers, Biophys. J. 76 (1999) 2142–2157. [32] Y. Lange, J. Ye, T.L. Steck, Activation of membrane cholesterol by displacement from phospholipids, J. Biol. Chem. 280 (2005) 36126–36131. [33] Megha, E. London, Ceramide selectively displaces cholesterol from ordered lipid domains (rafts): implications for lipid raft structure and function, J. Biol. Chem. 279 (2004) 9997–10004. [34] S. Nybond, Y.J.E. Bjorkqvist, B. Ramstedt, J.P. Slotte, Acyl chain length affects ceramide action on sterol/sphingomyelin-rich domains, Biochim. Biophys. Acta Biomembr. 1718 (2005) 61–66. [35] C.J. Yu, M. Alterman, R.T. Dobrowsky, Ceramide displaces cholesterol from lipid rafts and decreases the association of the cholesterol binding protein caveolin-1, J. Lipid Res. 46 (2005) 1678–1691. [36] A.M. Jimenez-Monreal, J. Villalain, F.J. Aranda, J.C. Gomez-Fernandez, The phase behavior of aqueous dispersions of unsaturated mixtures of diacylglycerols and phospholipids, Biochim. Biophys. Acta Biomembr. 1373 (1998) 209–219. [37] O. Berger, O. Edholm, F. Jahnig, Molecular dynamics simulations of a fluid bilayer of dipalmitoylphosphatidylcholine at full hydration, constant pressure, and constant temperature, Biophys. J. 72 (1997) 2002–2013. [38] W.F. van Gunsteren, P. Kruger, S.R. Billeter, A.E. Mark, A.A. Eising, W.R.P. Scott, P.H. Huneberg, I.G. Tironi, Biomolecular Simulation: The GROMOS96 Manual and User Guide, Biomos Hochschulverlag AG an der ETH Zurich, Groningen, Germany, 1996.

263

[39] B. Hess, C. Kutzner, D. van der Spoel, E. Lindahl, GROMACS 4: algorithms for highly efficient, load-balanced, and scalable molecular simulation, J. Chem. Theory Comput. 4 (2008) 435–447. [40] W.G. Hoover, Canonical dynamics: equilibrium phase-space distributions, Phys. Rev. A 31 (1985) 1695–1697. [41] M. Parinello, A. Rahman, Polymorphic transitions in single crystals: a new molecular dynamics method, J. Appl. Phys. 52 (1981) 7182. [42] U. Essmann, L. Perera, M.L. Berkowitz, T. Darden, H. Lee, L.G. Pedersen, A smooth particle mesh Ewald method, J. Chem. Phys. 103 (1995) 8577–8593. [43] A.W. Schuettelkopf, D.M.F. van Aalten, PRODRG — a tool for high-throughput crystallography of protein–ligand complexes, Acta Crystallogr. D60 (2004) 1355–1363. [44] W. Humphrey, A. Dalke, K. Schulten, VMD: visual molecular dynamics, J. Mol. Graph. 14 (1996) 33–38 27–38. [45] D. Van Der Spoel, E. Lindahl, B. Hess, G. Groenhof, A.E. Mark, H.J. Berendsen, GROMACS: fast, flexible, and free, J. Comput. Chem. 26 (2005) 1701–1718. [46] G. Bussi, D. Donadio, M. Parrinello, Canonical sampling through velocity rescaling, J. Chem. Phys. 126 (2007). [47] H.J.C. Berendsen, J.P.M. Postma, W.F. Vangunsteren, A. Dinola, J.R. Haak, Moleculardynamics with coupling to an external bath, J. Chem. Phys. 81 (1984) 3684–3690. [48] W.F. Bennett, J.L. MacCallum, D.P. Tieleman, Thermodynamic analysis of the effect of cholesterol on dipalmitoylphosphatidylcholine lipid membranes, J. Am. Chem. Soc. 131 (2009) 1972–1978. [49] D.P. Tieleman, S.J. Marrink, Lipids out of equilibrium: energetics of desorption and pore mediated flip-flop, J. Am. Chem. Soc. 128 (2006) 12462–12467. [50] J.S. Hub, B.L. de Groot, D. van der Spoel, g_wham—a free weighted histogram analysis implementation including robust error and autocorrelation estimates, J. Chem. Theory Comput. 6 (2010) 3713–3720. [51] M. Alwarawrah, J. Dai, J. Huang, A molecular view of the cholesterol condensing effect in DOPC lipid bilayers, J. Phys. Chem. B 114 (2010) 7516–7523. [52] S. Ollila, M.T. Hyvonen, I. Vattulainen, Polyunsaturation in lipid membranes: dynamic properties and lateral pressure profiles, J. Phys. Chem. B 111 (2007) 3139–3150. [53] A.M. Smondyrev, M.L. Berkowitz, Structure of dipalmitoylphosphatidylcholine/ cholesterol bilayer at low and high cholesterol concentrations: molecular dynamics simulation, Biophys. J. 77 (1999) 2075–2089. [54] L.S. Vermeer, B.L. de Groot, V. Reat, A. Milon, J. Czaplicki, Acyl chain order parameter profiles in phospholipid bilayers: computation from molecular dynamics simulations and comparison with 2H NMR experiments, Eur. Biophys. J. 36 (2007) 919–931. [55] D.L. Armstrong, D.B. Borchardt, R. Zidovetzki, Synergistic perturbation of phosphatidylcholine/sphingomyelin bilayers by diacylglycerol and cholesterol, Biochem. Biophys. Res. Commun. 296 (2002) 806–812. [56] D. Armstrong, R. Zidovetzki, Amplification of diacylglycerol activation of protein kinase C by cholesterol, Biophys. J. 94 (2008) 4700–4710. [57] J. Dai, M. Alwarawrah, J. Huang, Instability of cholesterol clusters in lipid bilayers and the cholesterol's umbrella effect, J. Phys. Chem. B 114 (2010) 840–848. [58] S.A. Pandit, D. Bostick, M.L. Berkowitz, Complexation of phosphatidylcholine lipids with cholesterol, Biophys. J. 86 (2004) 1345–1356. [59] S.A. Pandit, S.W. Chiu, E. Jakobsson, A. Grama, H.L. Scott, Cholesterol surrogates: a comparison of cholesterol and 16:0 ceramide in POPC bilayers, Biophys. J. 92 (2007) 920–927. [60] M.R. Ali, K.H. Cheng, J. Huang, Ceramide drives cholesterol out of the ordered lipid bilayer phase into the crystal phase in 1-palmitoyl-2-oleoyl-sn-glycero-3phosphocholine/cholesterol/ceramide ternary mixtures, Biochemistry 45 (2006) 12629–12638. [61] J. Huang, J.T. Buboltz, G.W. Feigenson, Maximum solubility of cholesterol in phosphatidylcholine and phosphatidylethanolamine bilayers, Biochim. Biophys. Acta 1417 (1999) 89–100. [62] A. Parker, K. Miles, K.H. Cheng, J. Huang, Lateral distribution of cholesterol in dioleoylphosphatidylcholine lipid bilayers: cholesterol–phospholipid interactions at high cholesterol limit, Biophys. J. 86 (2004) 1532–1544. [63] A. Bjorkbom, T. Rog, K. Kaszuba, M. Kurita, S. Yamaguchi, M. Lonnfors, T.K.M. Nyholm, I. Vattulainen, S. Katsumura, J.P. Slotte, Effect of sphingomyelin headgroup size on molecular properties and interactions with cholesterol, Biophys. J. 99 (2010) 3300–3308. [64] I. Artetxe, C. Sergelius, M. Kurita, S. Yamaguchi, S. Katsumura, J.P. Slotte, T. Maula, Effects of sphingomyelin headgroup size on interactions with ceramide, Biophys. J. 104 (2013) 604–612. [65] A.L. Kuo, C.G. Wade, Lipid lateral diffusion by pulsed nuclear magnetic resonance, Biochemistry 18 (1979) 2300–2308. [66] J.R. Sheats, H.M. McConnell, A photochemical technique for measuring lateral diffusion of spin-labeled phospholipids in membranes, Proc. Natl. Acad. Sci. U. S. A. 75 (1978) 4661–4663. [67] H.C. Gaede, K. Gawrisch, Lateral diffusion rates of lipid, water, and a hydrophobic drug in a multilamellar liposome, Biophys. J. 85 (2003) 1734–1740. [68] N. Kahya, Protein–protein and protein–lipid interactions in domain-assembly: lessons from giant unilamellar vesicles, Biochim. Biophys. Acta Biomembr. 1798 (2010) 1392–1398.