張濤，仇運廣，羅啟超，程曦，趙麗芬，嚴昕,4，彭浡，蔣華良,,＊，陽懷宇,＊

1華東理工大學藥學院，上海 200237

2中國科學院上海藥物研究所，新藥研究國家重點實驗室藥物發(fā)現(xiàn)與設計中心，上海 201203

3中國科學院大學，北京 100049

4上?？萍即髮W生命科學與技術學院，上海 201210

1 Introduction

A biological membrane is a selectively permeable barrier between the inside and outside environments of a cell in various living organisms that mediates signal transduction, cell-cell recognition and communication1. The cell envelope in gramnegative bacteria, such as Escherichia coli (E. coli), consist of two distinct membranes, specifically the outer and inner(cytoplasmic) membranes, separated by a cell wall2,3. The outer membrane is highly asymmetric with an inner phospholipid monolayer and an outer leaflet composed of lipopolysaccharides and complex macromolecules that can be classified into three structural components, namely, lipid A, the core oligosaccharide,and the O-antigen2,4. In contrast, the inner membrane is a classic symmetric phospholipid bilayer with numerous membrane proteins imbedded within it5,6. E. coli can survive in the gut, sea water, and sewage eff l uents7, and over a wide pH range in the presence of many ions, which is due to the high impermeability of its outer membrane and eff l ux-inf l ux transport systems in its inner membrane8,9.

Ions, especially divalent cations, directly associate with membrane lipids and thus influence the structural and functional properties of the membranes, for example, by neutralizing the phospholipid head groups10, participating in lipid-lipid packing11,12, and membrane fusion13,14. Ca2+and Mg2+ions are the main divalent cations in living cells15,16. Many publications suggest that these two cations are involved in the structural integrity and functioning of the E. coli outer membrane17-20.Gangola et al. reported that the level of intracellular Ca2+in E.coli is 100 to 1000-fold higher than that of free Ca2+in the cytoplasm21. Recent studies show that free cytosolic Ca2+levels in various bacteria are quite low (100-300 nmol·L-1)22. Mg2+ion levels in the cytoplasm were reported to be in the range of 200 μmol·L-1- 1 mmol·L-123,24. Therefore, the concentrations of Ca2+and Mg2+ions bound to the E. coli inner membrane vary over a wide range.

Phosphatidylethanolamine and phosphatidylglycerol are the major lipid components of the E. coli inner membrane. A mixed lipid bilayer composed of 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine (POPE) and 1-palmitoyl-2-oleoyl-snglycero-3-phosphoglycerol (POPG) in a 3 : 1 ratio (mol/mol) is a widely accepted model of the bacterial inner membrane used in research25-27. To our knowledge, recent studies have mainly focused on the interactions of Ca2+and Mg2+ions with POPG-containing lipid bilayers at a specific concentration25,27-29.However, the quantitative effects of Ca2+and Mg2+ions on the structure of POPE/POPG (3 : 1, mol/mol) bilayers over a wide range of concentrations are still unknown and should be investigated.

All-atom molecular dynamics (AA-MD) simulations have been used as an invaluable tool to provide key insights into the atomic-level details of ion-lipid interactions, the dyncamic conformations of pepetides and the vibrational spectra of ionic liquids that cannot be obtained by other techniques28,30-34.Dynamic light scattering (DLS) and zeta potential measurements are typically used to measure changes in the macroscopic properties of vesicles in response to other factors, such as ions and small molecules35-39. Here, the distinct differences in the binding affinities of Ca2+and Mg2+ions to POPE/POPG bilayers and their effects on the bilayers at different concentrations were determined by a combination of DLS and zeta potential measurements and AA-MD simulations.

2 Material and methods

2.1 Chemical reagents

High-purity (＞ 99%, HPLC grade) POPE and POPG in chloroform solutions (25 mg·mL-1) were purchased from Avanti Polar Lipids, Inc. (Alabaster, AL, USA). Both the POPE and POPG solutions were stored at -20 °C. The structures of POPE and POPG are shown in Fig. 1. 4-(2-Hydroxyethyl)-1-piperazineethanesulfonic acid (Hepes, 99.5%), CaCl2·2H2O(＞ 99.5%) and MgCl2·6H2O (≥ 99.5%) were purchased from Sigma-Aldrich China, Inc. NaCl (analytical-reagent grade),ethylenediaminetetraacetic acid (EDTA, analytical-reagent grade), CHCl3(HPLC grade) and CH3OH (HPLC grade) were purchased from Ourchem (Shanghai, China). All chemicals were used without further purification.

Fig. 1 Chemical structures and atom labeling of POPE and POPG.

2.2 Liposome preparation and characterization

A modified version of a previously reported protocol was employed to prepare the vesicles used for the DLS and zeta potential measurements40. The lipid solutions were added to CHCl3/CH3OH (2 : 1, vol/vol) in a round-bottom fl ask to obtain the desired composition of the POPE:POPG lipids (3 : 1,mol/mol). The solvents were evaporated in a gentle gaseous nitrogen stream and the sample were subsequently dried under vacuum for approximately four hours to remove the residual solvent. The samples were hydrated with a CaCl2or MgCl2buffer with a concentration of 0, 0.1, 0.5, 1, 3, 5, 10, 20, 50 mmol·L-1, or 100 mmol·L-1. The final concentration of the liposomes was approximately 2 mg·mL-1. Then, the liposomes were vortexed at 50 °C higher than the main phase transition temperature of the lipids until the mixture was optically homogeneous. The liposomes were formed by extrusion through a 100 nm membrane filter (Avanti Polar Lipids, Inc., Alabaster,AL, USA) to obtain large unilamellar vesicles (LUVs). Then, the size and zeta potential of the liposomes were determined by DLS and zeta potential measurements (Nano ZS 90, Malvern, UK),respectively.

2.3 Simulation systems

To investigate the detailed effects of Ca2+and Mg2+ions on the membrane at the atomic level, a lipid bilayer composed of 162 POPE and 54 POPG molecules (POPE/POPG, 3 : 1, mol/mol)(Fig. 2a) was constructed using the CHARMM-GUI program41.Seven systems, including one pure membrane system(membrane with 0 mmol·L-1Ca2+or Mg2+ions) (Fig. 2b) and six membrane systems with 5, 50 and 100 mmol·L-1Ca2+(Fig.2c) or Mg2+(Fig. 2d) ions were designed. Accordingly, 1, 10,and 20 Ca2+or Mg2+ions were added to six pure membranes.The temperature of each system was maintained at 300 K. Each of the systems was solvated with TIP3P water and neutralized with 0.15 mol·L-1NaCl.

2.4 MD simulation details and data processing

MD simulations in the isobaric-isothermal ensemble were performed using the GROMACS 5.1.4 package, with periodic boundary conditions42. The CHARMM 36 force field was used43.To relieve unfavorable contacts, energy minimizations were performed, followed by equilibration runs of 10 ns. The vrescale method44was used with a coupling time of 0.1 ps to maintain the temperature of the system, and the Nose-Hoover method45was subsequently employed with a coupling time of 0.5 ps for the production runs. The pressure was maintained at 1.0 × 105Pa using the Parrinello-Rahman method with τ = 1.0 ps and a compressibility of 4.5 × 10-10Pa-146. SETTLE47and LINCS48constraints were applied to the covalent bonds involving hydrogen atoms in water and other molecules,respectively. The time step was set to 2 fs, and the long-range electrostatic interactions were calculated using the particle mesh Ewald (PME) algorithm49. A cutoff value of 1.2 nm was used for both the electrostatic and van der Waals interaction calculations. The simulations of all the systems were performed for 300 ns, and the MD trajectories of last 100 ns were used for analysis. The area per lipid and membrane thickness were calculated using VMD 1.9.1 MEMBPLUGIN50. The gmx_minidist method was used to calculate the minimum distance between the ions and lipids. The radial distribution functions of the cations were calculated using the gmx_rdf program.

Fig. 2 Initial snapshots of simulated systems. POPE and POPG are represented as cyan and orange line, respectively (a). The cells are shown as black boxes, and Na+ (b), Ca2+ (c), and Mg2+ (d) ions are shown as grey, blue, and red spheres, respectively (for clarity, the Clions and water molecules are not shown).

All the data were statistically analyzed and plotted using the Origin 9.0 software (Origin Lab Corporation, Northampton,USA) or MATLAB software (MathWorks, Santa Clara, CA,USA). All the statistical data are presented as the mean ±standard deviation (SD).

3 Results and discussion

3.1 DLS and zeta potential measurements

First, both the Ca2+and Mg2+ions were found to induce changes in the macroscopic properties of the POPE/POPG vesicles. During the preparation of the POPE/POPG liposomes,lipid aggregates and precipitation were observed in the solutions containing 5-100 mmol·L-1CaCl2, which is consistent with another study that reported that Ca2+ions can induce the aggregation or fusion of negatively charged lipid vesicles35.Meanwhile, fewer lipid aggregates and less precipitation were observed in the solutions containing 5-100 mmol·L-1MgCl2.

To explain these phenomena and evaluate the quantitative effects of the Ca2+and Mg2+ions on the POPE/POPG liposomes,the size distributions of the LUVs were determined by DLS measurements in the ion concentration range of 0-100 mmol·L-1. To present a clear picture of the distinct differences between the effects of the Ca2+and Mg2+ions on the liposomes,the distribution profiles of the liposome hydrodynamic diameters were plotted for the two ions at the typical concentrations of 0,1, 5, and 100 mmol·L-1. As shown in Fig. 3a, the liposome hydrodynamic diameter exhibited a unimodal distribution in the 0, 1 mmol·L-1Ca2+and 1 mmol·L-1Mg2+ion solutions, with average values of (164.1 ± 91.1), (192.8 ± 84.1) and (200.2 ±103.3) nm, respectively, indicating that the liposomes were homogeneous and monodisperse under these conditions. When the Ca2+ion concentration was increased to 5 and 100 mmol·L-1,the distribution profiles shifted to the right and became bimodal with average liposome diameters of (2095.8 ± 1080.9) and(2414.9 ± 1524.9) nm, respectively, demonstrating that lipid aggregation or liposome fusion of the unilamellar liposomes events occurred, which is consistent with previous research35.When the liposomes were in a 5 mmol·L-1Mg2+ion solution, the distribution profile of the liposome hydrodynamic diameter also shifted to the right, but it became broader than that obtained in the solutions containing 5 mmol·L-1Ca2+ions. In this case, the average liposome hydrodynamic diameter was (1122.5 ±1310.7) nm. When the Mg2+ion concentration increased to 100 mmol·L-1, a multimodal distribution profile was observed for the liposome hydrodynamic diameter, which averaged (1466.1 ±1356.7) nm (Fig. 3b). These results suggest that both the Ca2+and Mg2+ions slightly affected the POPE/POPG liposomes at relatively low concentrations (1 mmol·L-1), whereas the Ca2+ions had more significant effects on the liposomes than the Mg2+ions at relatively high ion concentrations (5-100 mmol·L-1).

Fig. 3 Liposome size distributions and zeta potential. (a, b) Hydrodynamic diameters of extruded LUVs composed of POPE/POPG (3 : 1, mol/mol)in Ca2+ or Mg2+ solutions with different concentrations. (c) Zeta potential of POPE/POPG (3 : 1, mol/mol) LUVs measured as a function of Ca2+ and Mg2+ concentrations.

To further evaluate vesicle clustering, the zeta potential of the POPE/POPG liposomes were measured in both Ca2+and Mg2+solutions with different concentrations. Clearly, the zeta potential profile of the liposomes obtained in the presence of Ca2+ions was higher than that obtained in the presence of Mg2+ions (Fig. 3c). The initial zeta potential of the POPE/POPG liposomes in the absence of divalent cations was (-23.7 ± 1.5)mV, which is roughly consistent with that of 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine/1-palmitoyl-2-oleoyl-snglycero-3-phospho-L-serine (POPC/POPG) (80 : 20, mol/mol)liposomes, as reported by Marco M. Domingues et al.36. As shown in Fig. 3c, the zeta potential increased as the concentrations of both of the ions increased. The zeta potential became positive when the Ca2+ion concentration was higher than ～58 mmol·L-1, while it became positive when the concentration of the Mg2+ions was higher than ～85 mmol·L-1.These results provided direct evidence that both the Ca2+and Mg2+ions had overcharging effects on the negatively charged POPE/POPG liposomes; similar phenomena were observed in the POPC/POPS (1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-L-serine)-Ca2+system by Melcrová et al.35. However, the effects of the Ca2+ions on the liposomes were more significant than those of the Mg2+ions. In addition, based on the theory reported by Melcrová et al., electrostatic repulsion between the positively charged vesicles might explain the vesicle fusion events caused by the two ions in the DLS experiments35.

3.2 Molecular dynamics simulations

To obtain atomic-level details of the ion-lipid interactions,AA-MD simulations were performed to investigate both Ca2+-and Mg2+-POPE/POPG bilayer interactions at various ion concentrations.

3.2.1 Cation-lipid interactions

First, to determine the difference in the adsorption abilities of Ca2+and Mg2+ions on the POPE/POPG bilayers, the number of lipid-ion contacts were calculated at different concentrations.When the distance between one cation and an oxygen atom of a lipid head group was within 0.42 nm, the cation was considered to contact the lipid. Fig. 4 shows the time evolution of the number of lipid-ion contacts. Clearly, all the Ca2+ions irreversibly adsorbed on the membrane when the simulation time was longer than 100 ns at concentrations of 5, 50 and 100 mmol·L-1. In contrast, the Mg2+ions exhibited adsorption/desorption behavior on the membrane during the 300 ns simulations at these ion concentrations. The difference in the adsorption behaviors of Ca2+and Mg2+on the lipid bilayer indicates that the number of Mg2+ions in the aqueous phase was larger than that of the Ca2+ions, while the mean lifetime of the Ca2+-lipid complex was longer than that of the Mg2+-lipid complex at the same concentration. Under equilibrium conditions at all three concentrations, 1 calcium ion was adsorbed per 4 lipids molecules, and 1 magnesium ion was adsorbed per 2 lipids molecules in the POPE/POPG bilayer,which is within the range of 1-4 lipids molecules per cation reported in earlier MD studies51,52. Thus, these results provide direct evidence that the binding affinity of Ca2+ions to POPE/POPG bilayers is much stronger than that of Mg2+ions at various concentrations, which might explain why the zeta potential became positive at a lower ion concentration in the Ca2+solution than that in the Mg2+solution.

Fig. 4 Time evolution of the fractions of bound Ca2+ (a) and Mg2+ (b)ions in a POPE/POPG (3 : 1, mol : mol) bilayer at various concentrations.

The binding of Ca2+and Mg2+ions by individual atomic groups in POPE and POPG was quantified by calculating the first coordination numbers, i.e., the average numbers of cations in the first coordination shell, of specific groups in the pure membrane and membranes containing 5 and 100 mmol·L-1ions from the radial distribution functions. A cutoff of 0.42 nm was employed for both the Ca2+- and Mg2+-phosphorous atom contacts, while the cutoff for the contacts of these two ions with the oxygen atoms of the carbonyl and hydroxyl groups was 0.3 nm. The average numbers of the considered groups in the first coordination shells of Ca2+and Mg2+ions adsorbed to the lipid bilayers are listed in Table 1. The average numbers of the phosphate groups of POPE and POPG in the first coordination shell of an adsorbed Ca2+ion were 0.018 and 0.037 in the 5 mmol·L-1Ca2+ion solutions, respectively. The average numbers of the phosphate groups of POPE and POPG in the first coordination shell of an adsorbed Mg2+ion were 0.003 and 0.003 in the 5 mmol·L-1Mg2+ion solutions, respectively. In contrast,the average number of carbonyl oxygen atoms of POPE and POPG and hydroxyl oxygen atoms of POPG in the first coordination shell of an adsorbed Ca2+ion or Mg2+ion was 0 in the 5 mmol·L-1Ca2+or 5 mmol·L-1Mg2+ion solutions. These results demonstrate that both the Ca2+and Mg2+ions preferred to bind to the phosphate groups of POPE and POPG at this concentration. When the Ca2+ion concentration was increased to 100 mmol·L-1, the coordination numbers of the phosphate groups of POPE and POPG were almost twelve- to twenty-fold higher than those obtained at 5 mmol·L-1, while the coordinationnumbers of the carbonyl oxygen atoms of POPE and POPG and hydroxyl oxygen atoms of POPG increased slightly. For Mg2+,the coordination numbers of the phosphate groups of POPE and POPG were almost twenty- to thirty-fold higher than those obtained at 5 mmol·L-1, and the coordination numbers of the carbonyl oxygen atoms of POPE and POPG and hydroxyl oxygen atoms of POPG varied negligibly. These results demonstrate that the main binding site of the Ca2+and Mg2+ions on POPE and POPG was the phosphate group at both 5 mmol·L-1and 100 mmol·L-1, which is consistent with other reports29.Therefore, these results might explain the overcharging effects of both the Ca2+and Mg2+ions on the POPE/POPG liposomes at the atomic level, because the electronegative phosphate groups could be gradually saturated by the cations as their concentrations increased.

Table 1 The average numbers of the considered groups in the first coordination shell of an adsorbed Ca2+ or Mg2+ ion to lipid bilayers.

3.2.2 Area per lipid and membrane thickness

Strong binding of cations might lead to the lateral compression of the lipid bilayers29,35. To verify this assumption,the area-per-lipid (Apl) and bilayer thickness (h = Dp-p) were calculated. Here, Aplwas calculated by dividing the lateral surface area of the simulation box by the number of lipids in a single layer, and Dp-pwas def i ned as the distance between the centers of mass of the phosphorus atoms in the top and bottom leaflets along the z-axis. The results for the membranes with different concentrations of Ca2+and Mg2+ions are listed in Table 2.The initial Aplvalue of the pure membrane was (0.56 ± 0.01)nm2, which is similar to the values obtained from previous MD simulations and to experimental results29,53. The Ca2+ions had a negligible effect on the Aplof the lipid bilayer at a concentration of 5 mmol·L-1(low concentration), while the Aplvalue decreased by 8.9% and 14.3% at concentrations of 50 and 100 mmol·L-1(high concentrations), respectively. In contrast, the Aplvalue of lipid bilayer hardly varied in the presence of Mg2+ions in the range of 5 to 100 mmol·L-1. These results demonstrate that the Ca2+ions laterally compressed the lipid bilayer at high concentrations, while the Mg2+ions did not compress the lipid bilayer at either low or high concentrations, which is consistent with earlier MD studies29. Compared to the pure membrane, the membrane thickness changed negligibly in the 5 mmol·L-1Ca2+ion solutions, but it increased by 6.1% and 10.4% in the 50 and 100 mmol·L-1Ca2+ion solutions, respectively. For the membranes in the Mg2+ion solutions, the thickness varied negligibly in the range of 5-100 mmol·L-1(Table 2). These results indicate that Apldecreased, while h increased with the increasing Ca2+ion concentration. Based on previous MD simulations, this phenomenon is mainly because the head groups of the lipids were pushed toward the water phase when the membrane was laterally compressed, resulting in a higher membrane thickness29. For the Mg2+ion, the low binding affinity to and adsorption/desorption behavior on the lipid bilayer might explain its slight effects on the Aplvalue and membrane thickness.

3.2.3 Atom location

To further verify the membrane thickness variations in the Ca2+and Mg2+ion solutions, the density profiles of the membrane components in the pure membrane and membranes containing 5 and 100 mmol·L-1Ca2+or Mg2+ions were calculated along the bilayer normal. The density profiles of the phosphorous and O22(Sn-2 carbonyl oxygen) atoms of POPE and POPG in the five systems are plotted in Fig. 5. In the pure membrane, the phosphorous and O22atoms of POPE were distributed in the ranges of 0.92-3.13 nm and 0.43-2.84 nm,respectively, with peak values of 2.17 and 1.69 nm, respectively.These two atoms of POPG were located in the ranges of 0.92-3.13 nm and 0.43-2.84 nm, respectively, with peak values of 2.17 and 1.78 nm, respectively. When the membrane was in a 5 mmol·L-1Ca2+ion solution, the phosphorous and O22atoms of POPE were distributed in the ranges of 0.93-3.18 nm and 0.44-2.89 nm, respectively, with peak values of 2.20 and 1.71 nm,respectively. These two atoms of POPG were located in the ranges of 1.13-3.18 nm and 0.54-2.79 nm, respectively, with peak values of 2.20 and 1.71 nm, respectively. When the membrane was in a 5 mmol·L-1Mg2+ion solution, thephosphorous and O22atoms of POPE were distributed in the ranges of 1.06-2.99 nm and 0.51-2.72 nm, respectively, with peak values of 1.98 nm and 1.61 nm, respectively. These two atoms of POPG were located in the ranges of 0.88-2.99 nm and 0.41-2.72 nm, respectively, with peak values of 2.20 and 1.71 nm, respectively. As the Ca2+ion concentration was increased to 100 mmol·L-1, the phosphorous and O22atoms of POPE were distributed in the ranges of 1.18-3.32 nm and 0.62-2.98 nm,respectively, with peak values of 2.42 and 1.97 nm, respectively.These two atoms of POPG were located in the ranges of 1.18-3.21 nm and 0.73-2.87 nm, respectively, with peak values of 2.53 and 1.97 nm, respectively. When the membrane was in a 100 mmol·L-1Mg2+ion solution, the phosphorous and O22atoms of POPE were distributed in the ranges of 0.91-3.11 nm and 0.43-2.83 nm, respectively, with peak values of 2.06 and 1.68 nm, respectively. These two atoms of POPG were located in the ranges of 1.10-3.21 nm and 0.53-2.83 nm, respectively, with peak values of 2.16 and 1.77 nm, respectively. Obviously, the atoms of POPE and POPG in the 100 mmol·L-1Ca2+-containing membrane shifted more toward the water phase than those in the other four systems, which is consistent with the calculated membrane thicknesses; the membranes containing high concentrations of Ca2+were thicker than the pure and Mg2+-containing membranes.

Table 2 Average area per lipid (Apl) and average membrane thicknesses (h) of the studied systems.

Fig. 5 Density profiles of the phosphorous and O22 (Sn-2 carbonyl oxygen) atoms of POPE (top) and POPG (bottom) in the pure membrane and membranes containing 5 and 100 mmol·L-1 Ca2+ (a, c), or Mg2+ (b, d).

3.2.4 Head group orientation

The interactions between the cations and the head groups of the lipids affect the head group angles, as demonstrated by many MD simulation studies28,31,54. Here, the effects of the Ca2+and Mg2+ions on the head group angles of POPE and POPG in the pure membrane and membranes containing 5 mmol·L-1(low concentration) and 100 mmol·L-1(high concentration) ions were investigated. For POPE, the orientation of the head group was def i ned as the tilt angle (θ) between the P→N vector (from the phosphorus atom to the nitrogen atom of POPE) and the outward normal of the membrane (Fig. 6a). For POPG, the orientation of the head group was def i ned as the tilt angle (Φ) between the P→C13 vector (from the phosphorus atom to C13 of POPG) and the outward normal of the membrane (Fig. 6c). As shown in Fig.6, the peaks in the θ probability distributions of POPE for the pure membrane, 5 mmol·L-1Ca2+-containing and 100 mmol·L-1Ca2+-containing membranes were observed at 87.7°, 87.2° and 88.8°, respectively, with average θ values of 87.5° ± 52.7°, 87.8° ±52.7°, and 88.0° ± 52.7°. The peaks in the θ probability distributions for the membranes containing 5 and 100 mmol·L-1Mg2+were observed at 87.7° and 87.2°, respectively, with average θ values of 88.0° ± 52.7° and 87.9° ± 52.7°, respectively.For POPG, the peaks in the Φ probability distributions for the pure, 5 mmol·L-1Ca2+- containing and 100 mmol·L-1Ca2+-containing membranes were observed at 87.8°, 88.4° and 87.4°,respectively, with average Φ values of 87.9° ± 52.7°, 88.3° ±52.7° and 87.6° ± 52.7°, respectively. The peaks in the Φ probability distributions for the membranes containing 5 and 100 mmol·L-1Mg2+were observed at 87.0° and 89.2°, respectively,with average Φ values of 87.7° ± 52.7° and 87.7° ± 52.7°. These results indicate that the head groups of POPE were slightly less perpendicular to the membrane surface at the high concentration of Ca2+ions than at the low concentration or in the absence of these ions. However, the presence of Mg2+ions had a negligible influence on the orientation of POPE at both the low and high concentrations. In contrast, POPG was slightly less perpendicular to the membrane surface at the low concentration of Ca2+ions than at the high concentration or in the absence of these ions,while the Mg2+ions had almost no effect on the orientation of POPG at both the low and high concentrations. Hence, the Ca2+ions affected the orientation of membrane lipids, while the Mg2+ions had negligible effects on this property of the membrane lipids. The calculated average tilt angles of POPE and POPG in our simulation systems are larger than those values, ranging from 66.9° ± 26.8° to 76.2° ± 27.4°, reported by Tsai et al.29. These differences might be attributed to several factors. First, the number of lipids in these systems is larger than that in the previous study. Second, the wide range of ion concentrations in this study differs from the concentration used in the previous study. Third, the simulation time and calculation software could explain the differences between the results of this study and the previous report. In contrast, the simulation results in this work are roughly consistent with the tilt angles of POPS in POPC/POPS simulation systems containing 100 mmol·L-1Ca2+35.

Fig. 6 Tilt-angle distributions of the P-N vector in POPE (top) and P-C13 vector in POPG (bottom) with respect to the z-axis for the pure membrane, and membranes containing 5 and 100 mmol·L-1 Ca2+ (a, c) or Mg2+ (b, d).

4 Conclusions

In this work, the quantitative effects of Ca2+and Mg2+ions on the structure of a POPE/POPG bilayer at different concentrations were determined by DLS, zeta potential measurements and AAMD simulations, and the following conclusions were made.First, the Ca2+ions had more severe effects on the POPE/POPG liposomes than the Mg2+ions did at the macroscopic scale.Second, both the Ca2+and Mg2+ions had overcharging effects on the negatively charged POPE/POPG liposomes. Third, the binding affinity of the Ca2+ions to the POPE/POPG bilayer was much stronger than that of the Mg2+ions, and the main binding site of the Ca2+and Mg2+ions on POPE and POPG was the phosphate group. Fourth, the Ca2+ions laterally compressed the lipid bilayer at high concentration, while the Mg2+ion exhibited no compression effects at either low or high concentrations. In addition, the membranes in the solutions with Ca2+ion concentrations were thicker than the pure membrane and the membranes in the Mg2+ion solutions. This result is consistent with the observation that the phosphorous and carbonyl oxygen atoms of POPE and POPG in the membrane containing a high concentration of Ca2+ions were shifted toward the water phase relative to those in the pure membrane and the Mg2+-containing membranes. Additionally, these two ions had different influences on the orientation of the lipid head groups at different concentrations. In conclusion, these findings give us valuable insight into the effects of divalent cations and other small molecules on the dynamics and structure of POPE/POPG bilayers.