Two-Dimensional Potentials of Mean Force of Nile ... - ACS Publications

Nov 18, 2015 - simulations we obtain positions and orientations of Nile Red ... to calculate the emission spectrum of Nile red as a function of its lo...
0 downloads 0 Views 5MB Size
Article pubs.acs.org/JCTC

Two-Dimensional Potentials of Mean Force of Nile Red in Intact and Damaged Model Bilayers. Application to Calculations of Fluorescence Spectra Gurpreet Singh, Adam C. Chamberlin, Hristina R. Zhekova, Sergei Y. Noskov, and D. Peter Tieleman* Department of Biological Sciences and Centre for Molecular Simulation, University of Calgary, 2500 University Drive N.W., Calgary, Alberta T2N1N4, Canada S Supporting Information *

ABSTRACT: Fluorescent dyes revolutionized and expanded our understanding of biological membranes. The interpretation of experimental fluorescence data in terms of membrane structure, however, requires detailed information about the molecular environment of the dyes. Nile red is a fluorescent molecule whose excitation and emission maxima depend on the polarity of the solvent. It is mainly used as a probe to study lipid microenvironments, for example in imaging the progression of damage to the myelin sheath in multiple sclerosis. In this study, we determine the position and orientation of Nile red in lipid bilayers by calculating twodimensional Potential of Mean Force (2D-PMF) profiles in a defect-free 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) bilayer and in damaged bilayers containing two mixtures of the oxidized lipid 1-palmitoyl-2-(9′-oxo-nonanoyl)-sn-glycero-3-phosphocholine and POPC. From 2D-PMF simulations we obtain positions and orientations of Nile Red corresponding to the minimum on the binding free energy surface in three different membrane environments with increasing amounts of water, mimicking damage in biological tissue. Using representative snapshots from the simulations, we use combined quantum mechanical/molecular mechanical (QM/MM) models to calculate the emission spectrum of Nile red as a function of its local solvation environment. The results of QM and QM/MM computations are in qualitative agreement with the experimentally observed shift in fluorescence for the dye moving from aqueous solution to the more hydrophobic environment of the lipid interiors. The range of the conformation dependent values of the computed absorption-emission spectra and the lack of solvent relaxation effects in the QM/MM calculations made it challenging to delineate specific differences between the intact and damaged bilayers.



tool for probing the lipid microenvironment.11 It is one of the most extensively used fluorescent probes for the characterization of the lipid or protein microenvironment. The dye is commonly used for staining intracellular lipid droplets for detection by fluorescence microscopy or flow cytofluorometry.12 A lipid bilayer microenvironment is complex. There are large gradients in density, polarity, and other physicochemical properties along the bilayer normal.13 This presents a challenge in relating experimentally observed spectral shifts to a change in membrane properties.14 Molecular simulation can be used to obtain the location and orientation of dye molecules and the lipid microenvironment surrounding a dye molecule, in atomistic details. A few such studies have been reported in the literature. Hoff et al. studied the distribution and orientation of the simple dye pyrene by simulations and solid state NMR and found good agreement between the two approaches.15 Vos et al. simulated a

INTRODUCTION The physicochemical properties of a biomembrane determine its biological function. Composition, phase state, polarity, level of hydration, and bilayer thickness affect biological activities such as transport of molecules across the bilayer or binding and insertion of proteins in the membrane.1,2 Therefore, monitoring these properties is crucial in understanding their biological function. Moreover, changes in lipid composition have been associated with pathophysiology of several diseases. Lipid peroxidation has been shown to be associated with Parkinson’s3 and Alzheimer’s disease,4 schizophrenia,5 atherosclerosis,6 and multiple sclerosis.7 Hence, tools such as solvatochromic dyes, that can probe changes in the lipid microenvironment, are also used in monitoring the progression of certain diseases.8,9 Solvatochromic dyes are dyes whose position of the excitation and emission maxima are dependent on the polarity of their environment. 10 Nile red, 9-diethylamino-5H-benzo[α]phenoxazine-5-one, is one such solvatochromic dye. In case of Nile red, a shift toward shorter wavelengths is observed with decreasing solvent polarity. Nile red has proven to be a useful © XXXX American Chemical Society

Received: June 3, 2015

A

DOI: 10.1021/acs.jctc.5b00520 J. Chem. Theory Comput. XXXX, XXX, XXX−XXX

Article

Journal of Chemical Theory and Computation

Figure 1. Ball and stick representation of 9-diethylamino-5H-benzo[α]phenoxazine-5-one (Nile red) and N,N-dimethylaniline (DMA). The dihedral ϕ1 and ϕ2 corresponds to dihedrals involving atoms C10−C1−C12−C13 and C1−C12−C13−C17, respectively. The vector v1 is shown with a yellow arrow.

transmembrane helix with the probe AEDANS.16 Gullapalli et al. studied 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI-C18(3)) in a 1,2-dipalmitoyl-sn-glycero-3phosphatidyl choline (DPPC) bilayer using equilibrium molecular dynamics simulations to investigate orientation, dynamics, and the effect of different charge states of the dye.17 Barucha-Kraszeweska et al. used MD simulations to identify the location and interactions of the commonly used dye 6-propionyl2-dimethylaminonaphthalene (PRODAN) and its derivative 2dimethylamino-6-lauroylnaphthalene (LAURDAN) with a dioleoylphosphatidylcholine (DOPC) lipid bilayer.18 The absorption and emission spectra of PRODAN in a DOPC bilayer were calculated by Cwiklik et al. combining MD simulations with time dependent density functional theory (TDDFT).19 Murugan et al. studied interactions of N-acetylaladanamide (NAAA) with a POPC lipid bilayer using simulations to assess structure and dynamics of the membrane-bound molecule.20 Computational studies dedicated to the orientation of pyrene and other probes in various lipids21,22 and to the fluorescence solvent relaxation effects in lipid systems containing LAURDAN23 have been published as well. In this study, we determine the depth and orientations of the complex Nile red molecule in a POPC bilayer and in two mixtures consisting of 1-palmitoyl-2-(9′-oxo-nonanoyl)-sn-glycero-3-phosphocholine (PoxnoPC) and POPC in 1:8 and 1:4 ratios. PoxnoPC is a stable oxidation product of POPC and a product of ozone mediated oxidation of lung surfactant extract.24,25 We develop an all-atom description of the dye that is compatible with OPLS and AMBER force fields, representing the ground state of Nile red. Using this model, we calculate a twodimensional potential of mean force (2D PMF) profile as a function of distance and orientation with respect to the membrane normal. Expanding upon the insight gained from the 2D PMFs in the membrane, we employ the Quantum Mechanical/Molecular Mechanical (QM/MM) method to compute absorption and emission spectra of Nile red in bilayer environment.

generated using Generalized Amber Force Field and the ACPYPE software.26 The dihedral parameters for ϕ1 and ϕ2 obtained after fitting were used in the MM topology of Nile red. For validation of the final Nile red topology, the chloroform to water transfer free energy was computed and compared with experimental values. The details are provided in the Supporting Information. PMF Profile Calculations. The simulation box consisted of a bilayer patch of 128 lipid molecules and 6096 water molecules. The mixed bilayers were prepared by randomly replacing 8 or 16 POPC molecules in each leaflet with PoxnoPC lipids. Periodic boundary conditions were applied in all three dimensions, with the bilayer normal along the Z dimension. The Z component of the distance between the center of masses of the bilayer and the Nile red molecule was defined as collective variable ζ1, while the angle between the vector v1 (Figure 1) and the Z axis was defined as ζ2. We computed the 1D PMF along ζ1 for the dye in the pure POPC bilayer, whereas 2D PMFs were computed for all three bilayers (pure POPC bilayer and in bilayers containing 1:8 or 1:4 PoxnoPC-POPC mixtures). For the 1D PMF we used 39 umbrella simulations uniformly distributed such that 0 ≤ ζ1 < 3.9 nm. The duration of each umbrella sampling window was 300 ns. The simulations were performed at 298 K. For computing 2D PMFs, we used a uniform grid of 1 Å, 5° and a range of 0 ≤ ζ1 < 3.8 nm, 0 ≤ ζ2 < 180° except in the case of the system with a 1:4 PoxnoPC-POPC bilayer where a range of 0 ≤ ζ1 < 2.5 nm was used. For each lipid bilayer up to 1368 simulations, each 16 ns long, were carried out (i.e., the 1D-PMF simulations were approximtely 20 times longer than the 2D-PMF ones). Further details of the setup are provided in the Supporting Information. The data obtained was used to compute 2D PMF profiles with a method described by Zhu et al. In brief, the negative log likelihood  was minimized using numerical optimization algorithms27 s



 (g1 , ..., gs) = −∑ Ng i i − i=1

COMPUTATIONAL METHODS Parametrization. Nile red has a diethyl amino group connected to a conjugated system via a C−N bond for which we did not have viable dihedral parameters. We primarily focused our parametrization effort on the dihedral connecting the diethyl amino group to the phenoxazine ring. We used N,N-dimethylaniline (DMA) to obtain dihedral profiles for the rotation around the N−C bond. The atoms used for defining ϕ1 and ϕ2 dihedrals are shown in Figure 1. The initial MM topology of Nile red was

M

∑ Ml ln l=1

Ml ∑i Nci ile gi

(1)

Here, s is the total number of independent simulations, M is the total number of bins, Ni and gi are the total number of samples and log of normalization factor for simulation i, respectively, and cil = e−wi(xl)/kBT is the exponential of the biasing potential evaluated at the center of bin l. The statistical inefficiency f i of an umbrella window i is computed as f i = 1 + 2τi where τi is the integrated autocorrelation time. The uncertainty in the computed PMFs relative to the PMF B

DOI: 10.1021/acs.jctc.5b00520 J. Chem. Theory Comput. XXXX, XXX, XXX−XXX

Article

Journal of Chemical Theory and Computation

Figure 2. Top-left panel: graph depicting QM profiles obtained by scanning ϕ1 dihedral at different levels of theory and basis sets. The MM energy profile for the GAFF force field is also shown. Top-right: Comparison of MM and QM potential energy profiles of ϕ1 rotation. The dihedral profiles obtained with the MP2/6-311++G** level of theory and the OPLS force field after fitting are shown. The potential energy surface of ϕ1, ϕ2 dihedrals obtained with the MP2/6-311++G** level of theory (left panel) and that obtained from OPLSFF after fitting are shown at bottom-left and bottom-right panels, respectively.

represented quantum mechanically and the surrounding solvent treated molecular mechanically, using UFF32 and the interaction of solvent with electronic field was accounted for using embedded charges. The structure of the dye was optimized in the field of the surrounding solvent, and the absorption peak for the lowest energy π−π* transition was computed with the solvent electronic field for the relaxed solute structure. The emission spectrum was computed for the excited state optimized structure of the dye, in the field of the surrounding molecular mechanical solvent. The molecular structures used for these calculations were all taken from the most representative structures of Nile red in a given bilayer. For determination of representative structures, all the frames where Nile red position and orientation were within 1 kJ/mol of the PMF minima were extracted and clustered using Daura’s clustering algorithm and RMSD as a distance metric.33 During the RMSD calculations, molecules were aligned using the least-squares fitting method but by only allowing rotation and translations in the XY plane. Nine randomly selected conformations from the most populated cluster along with the cluster center were used for optical spectra calculations. The selection of a fairly limited set of structures for evaluation of the emission spectra was dictated by the computational expense of the Gaussian calculations (QM/MM optimization of excited state geometry), which require large amounts of memory (∼48 GB on a single node for the serial version of Gaussian). The set of calculations necessary for evaluation of the emission energy of a single structure took ∼150 h.

tail in the bulk solution were obtained by using a bootstrapping method. The uncertainties were estimated as twice the standard deviation. The probability along ζ2 was further divided by sin(ζ2) to ensure proper normalization along ζ2. Optical Spectrum Calculations. The electronic structure and optical spectra of Nile red were calculated using Gaussian 0928 without minimization of the geometric structure except where explicitly stated. The molecular orbitals were optimized using the quadratically convergent SCF procedure29 with a convergence criterion of less than 10−7 au variation in the density using time-dependent density functional theory (TD-DFT).30 We chose the 6-31+G** basis set and CAM-B3LYP density functional for calculations of absorption and emission spectra after preliminary testing of different combinations of functionals and basis sets. Our test calculations were done on systems consisting of the Nile Red dye and increasing number of water molecules. Test calculations with implicit solvent models were performed as well, and the effect of solvent relaxation on the emission energies was investigated. See the Supporting Information for more details. The calculations of the optical spectra of the Nile Red dye embedded in the lipid bilayers proved rather challenging. Our attempts to decrease the computational costs by representing the environment exclusively as a collection of external point charges were unsuccessful. The best agreement with experiment was achieved when the surrounding lipid and water molecules were included in a QM/MM calculation, using the ONIOM model.31 The QM/MM calculations were carried out with the dye C

DOI: 10.1021/acs.jctc.5b00520 J. Chem. Theory Comput. XXXX, XXX, XXX−XXX

Article

Journal of Chemical Theory and Computation



RESULTS AND DISCUSSION Parametrization. The Nile red molecule is fairly rigid consisting of a ring moiety connected to a diethyl amino group via a C−N bond. It has been suggested that the ground state geometry of Nile red is planar, and rotation around the C−N bond occurs in the excited state.34 For building a MM model of the dye, we focused on the dihedral involving rotation around the C−N bond (ϕ1). Figure 2 shows the MM and QM energy profiles of rotation of ϕ1 for the DMA molecule. The barrier heights and geometries are different between the MP2 and DFT level of theory. At the MP2 level, the dihedral profiles are sensitive to the basis set. At the DFT level of theory, at ϕ1 = 0°, both the methyl groups are planar with respect to the aromatic ring (ϕ2 = 180°), whereas at MP2 level, the methyl carbon (C17) is 47° out of the plane (ϕ2 = 133°). The MM energy profile for GAFF is also shown. Here also, at ϕ1 = 0°, the minimum energy structure is a planar molecule. As the differences between QM and MM energy profiles were significant, we improved the dihedral parameters by fitting the MM energy profiles to the corresponding QM profiles. We used the energy profile obtained with MP2/6-311++G**, the highest level of theory used here, as a target profile for MM. We used a tabulated potential implementation in GROMACS to represent the dihedral potential for ϕ1, allowing an arbitrary function for ϕ1. The resulting MM energy profile is compared with the target QM profile in Figure 2. The use of the tabulated potential function resulted in a good fit to the target profile. We also modified the dihedral potential for ϕ2 to improve the representation of the underlying QM energy surface. The standard potential energy function Vdih = kdih(1−cos(nϕ − ϕs)) was used to represent the dihedral potential of ϕ2. The twodimensional QM energy profile and the corresponding MM profile are shown in Figure 2. The position and height of the minimum and maximum in MM profile matches the corresponding QM profile well. However, the agreement is poorer at the edges. It is likely that the agreement with QM potential energy surface could be further improved by modifying additional force field terms, but the Lennard-Jones parameters, bonds, and angles were taken from GAFF for consistency with the lipid parameters and were not treated as free parameters. The partial charges for DMA obtained by RESP fitting method were scaled to obtain a good agreement between experimental and computed ΔHvap. The final values for density and ΔHvap obtained from simulations at 298.15 K are 0.94 g/cm3 and 49.4 kJ/mol, respectively, and the corresponding experimental values are 0.95 g/cm3 and 49.7 kJ/mol, respectively. The partial charges for Nile red were also computed using the RESP fitting method. The resulting topology was used to calculate chloroform to water partitioning free energy using thermodynamic integration. The computed and experimental values at 277.15 K are 8 ± 2 and 12 kJ/mol, respectively. The partitioning free energy of the final MM model of Nile red is within 4 kJ of the experimentally observed value at 277.15 K indicating that the model is sufficiently accurate for PMF calculations. PMF Profiles. We have used three model membranes in this study. POPC is found abundantly in the animal cell membrane and is a neutral lipid. Figure 3 shows density profiles for the three bilayer systems, while Figure 4 shows snapshots of the local dye environment near its preferred location in the three different systems. The addition of oxidized lipids in the bilayer has been shown to increase the area per lipid and decrease the membrane thickness.35 As shown in Figure 3, we also observe increased

Figure 3. Density of water (black) and carbon atoms (green) along the membrane normal for pure POPC bilayer (solid lines) 1:8 PoxnoPCPOPC (dashed line) and 1:4 PoxnoPC-POPC (dotted lines). The density of water is labeled as W and is calculated from density of oxygen atoms of water molecule.

hydration of the headgroup region and a decrease in membrane thickness in mixed bilayers. Also, the level of hydration of the headgroup region increases with increasing PoxnoPC concentration. It is unlikely that a single MD simulation of a few hundreds nanoseconds will be able to sample all the relevant configurations of the dye inside a lipid environment. We used one- and twodimensional umbrella sampling to ensure equilibrated distributions of both depth inside the membrane and angle relative to the bilayer normal of the relatively rigid dye molecule. The PMF profile of Nile red along ζ1 in the POPC bilayer is shown in Figure 5 (left panel). The PMF profile indicates a strong preference for partitioning in the lipid bilayer as compared to the bulk water. The position of the PMF minimum is at ∼1.0 nm from the center of the bilayer. The orientation of the dye can be characterized to a certain extent by observing the frequency distribution of ζ2 from the biased simulation carried along ζ1, shown in Figure 5 (middle panel). The frequency distribution of ζ2 indicates that at ζ1 ∼ 0.5 or 1.5, the ζ2 is preferentially oriented between 160 and 180 deg, with no clear preference at ζ1 ∼ 1 nm. A reasonably accurate estimation of Nile red orientation in the region of interest, i.e. ζ1 ∼ 1 nm will require multiple independent observations of ζ2. The autocorrelation of ζ2 for each biased simulation along ζ1 can provide an indication of quality of sampling obtained along ζ2 and is shown in Figure 5 (right panel). The autocorrelation decay relatively fast (0.1−1.0 ns) for simulations with ζ1 > 2.8 nm, whereas for ζ1 < 2.0, the decay toward zero is slow, with the slowest decay of around 60 ns. As the autocorrelations in ζ2 are significantly large in simulations where 0 ≤ ζ1 ≤ 2, the estimation of preferred orientations is likely to be unreliable. The localization of the dye in the related lipid dioleoyl-snglycero-3-phosphocholine was studied by Mukherjee et al. using the parallex method. They concluded that the dye is located close to the water/lipid interface at a distance of 17.6 Å from the bilayer center and the position of the dye depends on the cholesterol concentration.36 In addition, the dye is motionally restricted. A direct comparison of this depth measurement is not straightforward as it is not clear, given the size of the dye D

DOI: 10.1021/acs.jctc.5b00520 J. Chem. Theory Comput. XXXX, XXX, XXX−XXX

Article

Journal of Chemical Theory and Computation

Figure 4. Representative snapshots of Nile red in POPC (right panel), 1:8 and 1:4 PoxnoPC-POPC bilayer (middle and left panels, respectively). The carbon atoms PoxnoPC lipids are colored blue.

Figure 5. PMF profile of Nile red as a function of distance from the POPC bilayer center (left). Frequency distribution of ζ2 (middle) and the time autocorrelations (right) of ζ2 for each each simulation performed along ζ1.

Figure 6. Two-dimensional PMF profile of Nile red as a function of distance from the bilayer center (ζ1 and angle with respect to the Z axis (ζ2 for the pure POPC bilayer (left), 1:8 PoxnoPC-POPC (center)) and 1:4 PoxnoPC-POPC bilayer (right). The contours are drawn every 2.0 kJ/mol.

molecule, what the distance experimentally represents exactly, but qualitatively the simulations do observe restricted motions and a location close to the water/lipid interface as the preferred depth.

In order to improve accuracy of our estimations, we computed 2D PMFs with ζ1 and ζ2 as the collective variables. The 2D PMF profiles of Nile red in POPC, 1:8 and 1:4 PoxnoPC-POPC bilayers are shown in Figure 6. Small variations in preferential E

DOI: 10.1021/acs.jctc.5b00520 J. Chem. Theory Comput. XXXX, XXX, XXX−XXX

Article

Journal of Chemical Theory and Computation

fluorescence spectra in three different environments, with increasing amounts of hydrophilic defects to investigate whether QM and QM/MM methods are able to reproduce shifts in spectra as a function of environment. Initial calculation of Nile red absorption spectra with just a water shell under a number of different conditions and methods (Table 1) shows that both mixed basis sets, despite the absence

orientations of the dye can be seen between the single component and two component bilayers. However, in all three bilayers, the position of the minima lies between 0.5 < ζ1 < 1.5 nm, 80 < ζ2 < 180 deg. The preferential orientation of the dye is such that the diethyl amino group is almost perpendicular to the bilayer normal. However, the orientation preference is not as strong as positional preference, especially for POPC, where the conformations with the diethyl amino group parallel to the bilayer normal are within thermal fluctuations of each other. The one-dimensional (1D) PMF profile along ζ1 can be obtained from the 2D PMF by integrating over ζ2. The 1D PMF profiles are compared in Figure 7. Overall the PMF profile

Table 1. Absorption (nm) of Nile Red in a 5 Å Water Shell Using QM/MM or Mixed Basis Sets with CAM-B3LYP parameter

value

gas phase QM − implicit solvent QM/MM (Nile red - QM/water - MM) mixed basis (Nile red - 6-31+G**/water - 3-21G) mixed Basis (Nile red -6-31+G**/water - 6-31G) full QM (Nile red - 6-31+G**/water - 6-3+1G**) experiment13 (water)

438 455 450 465 468 471 591

of polarization and diffuse functions on water, manage to capture most of the effect of the surrounding solvent on the absorption spectrum of Nile red. In the case of the QM/MM calculation the shift introduced by the molecular mechanical region in the probe is not as large as for the QM/QM cases; however, it does manage to capture a significant amount of the effect. Table 2 shows the corresponding absorption and emission energies taking into account explicitly the lipid environment. We observe a shift toward shorter wavelengths as the environment changes from pure water to a lipid/water bilayer. This is consistent with the experimentally determined blue-shifts for Nile red upon transition from a polar (water) to a nonpolar (heptane) solution (Table S1, Supporting Information). The absorption and emission spectra vary significantly between the different snapshots. However, the distinction between solvents in this case is not as well-defined. The variation between the spectra of the dye in the pure POPC bilayer and that of the 1:8 PoxnoPC:POPC bilayer is roughly 1.7 nm, which is smaller than the variation between different representative conformations for the dye in a given bilayer. When the results are averaged weighted by oscillator strength or after Boltzmann weighting of the structure, no notable distinction between these values was observed. The inconclusiveness of our optical spectra calculations is likely a combination of several factors: insufficient sampling (only 10 structures per bilayer type), lack of solvent relaxation, inherent problems in the QM/MM scheme (QM/ QM schemes afford better results but are not feasible in this case, see Table 1), choice of functional (CAM-B3LYP underestimates both the absorption and emission energies, see the Supporting Information). Our calculations of the Nile Red systems in implicitly included water show that the solvent relaxation leads to a bathochromic shift of the emission spectrum and improves the general agreement with experiment (Table S1, Supporting Information). Unfortunately, our present QM/MM results of the Nile red dye in a realistic environment of hundreds of lipid and water molecules do not account for solvent relaxation due to the computational cost of such calculations. In order to separate the contribution of the lipid oxidation state from the effect of the statistical fluctuations of the dye orientation within the lipid bilayer, we are currently trying to address the lack of relaxation effects in our QM/MM calculations and to expand considerably the test sets of PMF derived structures. We are also looking into other functionals (GGA and hybrid) that might improve the

Figure 7. PMF profile of Nile red along ζ1 for the pure POPC and 1:8 PoxnoPC-POPC bilayer systems. The PMF obtained from biased simulations only along ζ1 is labeled as popc(1D).

obtained from sampling only along ζ1 is similar to the 1D PMF profile obtained from sampling along both the collective variables. Between POPC and 1:8 PoxnoPC-POPC bilayers, the dye has a preference for the POPC bilayer. The position of the minimum in both systems is ∼1.0 nm. The width of the minimum is slightly different among the two systems with the POPC minimum slightly broader. Overall the 2D PMF profiles provide a better estimate of both the positional and orientational preferences of the dye as compared to the 1D PMFs due to the larger number of unbiased data points in 2D-PMF. Previous studies have demonstrated that obtaining statistical convergence of equilibrium properties for solute molecule embedding in a membrane is difficult, even with one-dimensional umbrella sampling along the bilayer normal.37 In the case of relatively rigid molecule such as Nile red, using both position and orientation as collective variable substantially improves the accuracy in PMF calculations. Optical Spectra Calculations. A blue shift in fluorescence spectra of Nile red has been reported for phosphatidylcholinetrioleoylglycerol mixtures as compared to phosphatidylcholine vesicles, dimyristoylphosphatidylcholine (DMPC)-cholesteryl linoleate microemulsions as compared to DMPC, and for the dye binding to very low or low density lipoproteins as compared to high density lipoproteins.13,14 Thus, Nile red is sensitive to the lipid environment and used for this purpose. We calculated F

DOI: 10.1021/acs.jctc.5b00520 J. Chem. Theory Comput. XXXX, XXX, XXX−XXX

Article

Journal of Chemical Theory and Computation

representative structures of Nile red in different bilayers. We observed that there is no significant difference among different lipid environments, as the variance between the conformations within the same bilayer is of similar magnitude as the variance among the three bilayer environments tested here. Improvements in the computational protocol used here, such as better sampling, better suited functionals, and inclusion of solvent relaxation effects, are necessary for the proper description of the Nile red fluorescence properties in the studied bilayers.

Table 2. QM-MM Calculation of the Absorption and Emission Spectra of Nile Red in Different Solvents POPC

average 1:4 PoxnoPC:POPC

average 1:8 PoxnoPC:POPC

average

absorption (nm)

emission (nm)

393.48 410.7 414.74 420.17 378.24 405.1 398.08 377.41 408.74 383 416.35 400.55 ± 46.71 388.47 419.97 385.49 407.43 411.89 411.25 380.85 385.63 414.20 376.28 376.84 396.21 ± 50.04 387.25 384.74 407.81 413.87 420.32 378.37 409.3 407.88 417.34 417.89 399.42 404.02 ± 43.71

476.21 480.77 476.68 476.33 471.62 475.25 473.96 468.82 477.84 468.97 479.93 475.13 ± 11.94 470.71 479.14 467.09 477.89 480.88 478.89 471.85 466.50 481.23 464.97 461.39 472.78 ± 21.39 469.46 467.37 472.44 473.68 481.25 466.43 473.29 476.52 477.59 476.6 473.68 473.48 ± 13.44



ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jctc.5b00520. Text describing additional methods and a range of validation results (PDF)



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This work was supported by grants from the Natural Sciences and Engineering Resarch Council (Canada) to S.Y.N. (RGPIN315019) and D.P.T. (RGPIN-238357). S.Y.N. is an Alberta Innovates Technology Futures New Faculty, Canadian Institute for Health Research New Investigator, and an Alberta Innovates Health Solutions (AIHS) Scholar. D.P.T. is an AIHS Scientist and Alberta Innovates Technology Futures Strategic Chair in (Bio)Molecular Simulation. H.Z. would like to thank Dr. Mauricio Chagas da Silva for his valuable input in the setup and analysis of QM and QM/MM calculations. Computations were performed on the West-Grid/Compute Canada facilities and the University of Calgary TNK cluster acquired with direct support from the Canada Foundation for Innovation.



calculated excitation energies. We would like to stress that the optical spectra results in the present work are preliminary, and we introduce them here mainly as a proof of concept.

REFERENCES

(1) Lee, A. G. How Lipids Affect the Activities of Integral Membrane Proteins. Biochim. Biophys. Acta, Biomembr. 2004, 1666, 62−87. (2) McIntosh, T. J.; Simon, S. A. Roles of Bilayer Material Properties in Function and Distribution of Membrane Proteins. Annu. Rev. Biophys. Biomol. Struct. 2006, 35, 177−198. (3) Dexter, D. T.; Carter, C. J.; Wells, F. R.; Javoy-Agid, F.; Agid, Y.; Lees, A.; Jenner, P.; Marsden, C. D. Basal Lipid Peroxidation in Substantia Nigra Is Increased in Parkinson’s Disease. J. Neurochem. 1989, 52, 381−389. (4) Markesbery, W. R.; Lovell, M. A. Four-Hydroxynonenal, a Product of Lipid Peroxidation, Is Increased in the Brain in Alzheimer’s Disease. Neurobiol. Aging 1998, 19, 33−36. (5) Dietrich-Muszalska, A.; Kontek, B. Lipid Peroxidation in Patients with Schizophrenia. Psychiatry Clin. Neurosci. 2010, 64, 469−475. (6) Esterbauer, H.; Wäg, G.; Puhl, H. Lipid Peroxidation and Its Role in Atherosclerosis. Br. Med. Bull. 1993, 49, 566−576. (7) Toshniwal, P. K.; Zarling, E. J. Evidence for Increased Lipid Peroxidation in Multiple Sclerosis. Neurochem. Res. 1992, 17, 205−207. (8) Stirling, D. P.; Cummins, K.; Mishra, M.; Teo, W.; Yong, V. W.; Stys, P. Toll-like Receptor 2-Mediated Alternative Activation of Microglia Is Protective after Spinal Cord Injury. Brain 2014, 137, 707−723. (9) McMillian, M. K.; Grant, E. R.; Zhong, Z.; Parker, J. B.; Li, L.; Zivin, R. A.; Burczynski, M. E.; Johnson, M. D. Nile Red Binding to HepG2



CONCLUSIONS We have developed a MM model of Nile red in the ground state and computed the 2D PMF profiles of the molecule as a function of its position and orientation in three bilayers differing in the amount of oxidized lipids. The addition of oxidized lipids increases the water density in the headgroup region of the bilayer and shifts the density maximum of lipid atoms toward the bilayer center, effectively decreasing the membrane thickness. The preferential position of the dye in all three systems is around 1.0 nm from the center of the bilayer. The dye is oriented in such a way that the largest principal axis is almost perpendicular to the bilayer normal; however, it could adopt other orientations with relative ease. In our investigation of the influence of the membrane upon the absorption and emission spectra of Nile red, we found that CAMB3LYP with the 6-31+G** basis set was sufficient to capture the qualitative behavior of Nile red in water. The absorption and emission values were obtained from QM/MM calculations on G

DOI: 10.1021/acs.jctc.5b00520 J. Chem. Theory Comput. XXXX, XXX, XXX−XXX

Article

Journal of Chemical Theory and Computation Cells: An Improved Assay for in Vitro Studies of Hepatosteatosis. In Vitro Mol. Toxicol. 2001, 14, 177−190. (10) Reichardt, C. Solvatochromic Dyes as Solvent Polarity Indicators. Chem. Rev. 1994, 94, 2319−2358. (11) Gao, F.; Mei, E.; Lim, M.; Hochstrasser, R. M. Probing Lipid Vesicles by Bimolecular Association and Dissociation Trajectories of Single Molecules. J. Am. Chem. Soc. 2006, 128, 4814−4822. (12) Greenspan, P.; Mayer, E. P.; Fowler, S. D. Nile Red: A Selective Fluorescent Stain for Intracellular Lipid Droplets. J. Cell Biol. 1985, 100, 965−973. (13) Greenspan, P.; Fowler, S. D. Spectrofluorometric Studies of the Lipid Probe, Nile Red. J. Lipid Res. 1985, 26, 781−789. (14) Sackett, D. L.; Wolff, J. Nile Red as a Polarity-Sensitive Fluorescent Probe of Hydrophobic Protein Surfaces. Anal. Biochem. 1987, 167, 228−234. (15) Hoff, B.; Strandberg, E.; Ulrich, A. S.; Tieleman, D. P.; Posten, C. 2H-NMR study and molecular dynamics simulation of the location, alignment and mobility of pyrene in POPC bilayers. Biophys. J. 2005, 88, 1818−1827. (16) Vos, W. L.; Schor, M.; Baumgaertner, A.; Tieleman, D. P.; Hemminga, M. A. Molecular dynamics simulations reveal that AEDANS is an inert fluorescent probe for the study of membrane proteins. Eur. Biophys. J. 2010, 39, 229−239. (17) Gullapalli, R. R.; Demirel, M. C.; Butler, P. J. Molecular Dynamics Simulations of DiI-C18(3) in a DPPC Lipid Bilayer. Phys. Chem. Chem. Phys. 2008, 10, 3548−3560. (18) Barucha-Kraszewska, J.; Kraszewski, S.; Jurkiewicz, P.; Ramseyer, C.; Hof, M. Numerical Studies of the Membrane Fluorescent Dyes Dynamics in Ground and Excited States. Biochim. Biophys. Acta, Biomembr. 2010, 1798, 1724−1734. (19) Cwiklik, L.; Aquino, A. J. A.; Vazdar, M.; Jurkiewicz, P.; Pittner, J.; Hof, M.; Lischka, H. Absorption and Fluorescence of PRODAN in Phospholipid Bilayers: A Combined Quantum Mechanics and Classical Molecular Dynamics Study. J. Phys. Chem. A 2011, 115, 11428−11437. (20) Murugan, N. A.; Apostolov, R.; Rinkevicius, Z.; Kongsted, J.; Lindahl, E.; Agren, H. Association Dynamics and Linear and Nonlinear Optical Properties of an N-Acetylaladanamide Probe in a POPC Membrane. J. Am. Chem. Soc. 2013, 135, 13590−13597. (21) Loura, L. M. S.; Martins do Canto, A. M. T.; Martins, J. Sensing hydration and behavior of pyrene in POPC and POPC/cholesterol bilayers: a molecular dynamics study. Biochim. Biophys. Acta, Biomembr. 2013, 1828, 1094−1101. (22) Loura, L. M. S.; Ramalho, J. P. P. Recent developments in molecular dynamics simulations of fluorescent membrane probes. Molecules 2011, 16, 5437−5452. (23) Jurkiewicz, P.; Cwiklik, L.; Jungwirth, P.; Hof, M. Lipid hydration and mobility: an interplay between fluorescence solvent relaxation experiments and molecular dynamics simulations. Biochimie 2012, 94, 26−32. (24) Uhlson, C.; Harrison, K.; Allen, C. B.; Ahmad, S.; White, C. W.; Murphy, R. C. Oxidized Phospholipids Derived from Ozone-Treated Lung Surfactant Extract Reduce Macrophage and Epithelial Cell Viability. Chem. Res. Toxicol. 2002, 15, 896−906. (25) Mahalka, A. K.; Maury, C. P. J.; Kinnunen, P. K. J. 1-Palmitoyl-2(9′-Oxononanoyl)-Sn-Glycero-3-Phosphocholine, an Oxidized Phospholipid, Accelerates Finnish Type Familial Gelsolin Amyloidosis in Vitro. Biochemistry 2011, 50, 4877−4889. (26) Sousa da Silva, A. W.; Vranken, W. F. ACPYPE - AnteChamber PYthon Parser interfacE. BMC Res. Notes 2012, 5, 367. (27) Zhu, F.; Hummer, G. Convergence and Error Estimation in Free Energy Calculations Using the Weighted Histogram Analysis Method. J. Comput. Chem. 2012, 33, 453−465. (28) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.; Scalmani, G.; Barone, V.; Mennucci, B.; Petersson, G. A.; Nakatsuji, H.; Caricato, M.; Li, X.; Hratchian, H. P.; Izmaylov, A. F.; Bloino, J.; Zheng, G.; Sonnenberg, J. L.; Hada, M.; Ehara, M.; Toyota, K.; Fukuda, R.; Hasegawa, J.; Ishida, M.; Nakajima, T.; Honda, Y.; Kitao, O.; Nakai, H.; Vreven, T.; Montgomery, J. A., Jr.; Peralta, J. E.; Ogliaro, F.; Bearpark, M.; Heyd, J. J.; Brothers, E.; Kudin,

K. N.; Staroverov, V. N.; Kobayashi, R.; Normand, J.; Raghavachari, K.; Rendell, A.; Burant, J. C.; Iyengar, S. S.; Tomasi, J.; Cossi, M.; Rega, N.; Millam, J. M.; Klene, M.; Knox, J. E.; Cross, J. B.; Bakken, V.; Adamo, C.; Jaramillo, J.; Gomperts, R.; Stratmann, R. E.; Yazyev, O.; Austin, A. J.; Cammi, R.; Pomelli, C.; Ochterski, J. W.; Martin, R. L.; Morokuma, K.; Zakrzewski, V. G.; Voth, G. A.; Salvador, P.; Dannenberg, J. J.; Dapprich, S.; Daniels, A. D.; Farkas, Ö .; Foresman, J. B.; Ortiz, J. V; Cioslowski, J.; Fox, D. J. Gaussian 09; Gaussian: 2009. (29) Bacskay, G. B. A Quadratically Convergent HartreeFock (QCSCF) Method. Application to Closed Shell Systems. Chem. Phys. 1981, 61, 385−404. (30) Scalmani, G.; Frisch, M. J.; Mennucci, B.; Tomasi, J.; Cammi, R.; Barone, V. Geometries and Properties of Excited States in the Gas Phase and in Solution: Theory and Application of a Time-Dependent Density Functional Theory Polarizable Continuum Model. J. Chem. Phys. 2006, 124, 94107. (31) Dapprich, S.; Komáromi, I.; Byun, K. S.; Morokuma, K.; Frisch, M. J. A New ONIOM Implementation in Gaussian 98. 1. The Calculation of Energies, Gradients and Vibrational Frequencies and Electric Field Derivatives. J. Mol. Struct.: THEOCHEM 1999, 461-462, 1−21. (32) Rappe, A. K.; Casewit, C. J.; Colwell, K. S.; Goddard, W. A.; Skiff, W. M. UFF, a Full Periodic Table Force Field for Molecular Mechanics and Molecular Dynamics Simulations. J. Am. Chem. Soc. 1992, 114, 10024−10035. (33) Daura, X.; Gademann, K.; Jaun, B.; Seebach, D.; Van Gunsteren, W. F.; Mark, A. E. Peptide Folding: When Simulation Meets Experiment. Angew. Chem., Int. Ed. 1999, 38, 236−240. (34) Golini, C. M.; Williams, B. W.; Foresman, J. B. Further Solvatochromic, Thermochromic, and Theoretical Studies on Nile Red. J. Fluoresc. 1998, 8, 395−404. (35) Wong-Ekkabut, J.; Xu, Z.; Triampo, W.; Tang, I.-M.; Tieleman, D. P.; Monticelli, L. Effect of Lipid Peroxidation on the Properties of Lipid Bilayers: A Molecular Dynamics Study. Biophys. J. 2007, 93, 4225−4236. (36) Mukherjee, S.; Raghuraman, H.; Chattopadhyay, A. Membrane Localization and Dynamics of Nile Red: Effect of Cholesterol. Biochim. Biophys. Acta, Biomembr. 2007, 1768, 59−66. (37) Neale, C.; Bennett, W. F. D.; Tieleman, D. P.; Pomès, R. Statistical Convergence of Equilibrium Properties in Simulations of Molecular Solutes Embedded in Lipid Bilayers. J. Chem. Theory Comput. 2011, 7, 4175−4188.

H

DOI: 10.1021/acs.jctc.5b00520 J. Chem. Theory Comput. XXXX, XXX, XXX−XXX