Sterol Binding and Membrane Lipid Attachment to the Osh4 Protein of

Oct 6, 2010 - ... sterol binding pocket, and were further stabilized by water-mediated interactions with polar residues at the bottom of the binding p...
0 downloads 0 Views 3MB Size
13562

J. Phys. Chem. B 2010, 114, 13562–13573

Sterol Binding and Membrane Lipid Attachment to the Osh4 Protein of Yeast Brent Rogaski, Joseph B. Lim, and Jeffery B. Klauda* Department of Chemical and Biomolecular Engineering, UniVersity of Maryland, College Park, Maryland 20742, United States ReceiVed: July 23, 2010; ReVised Manuscript ReceiVed: September 17, 2010

Osh4 is an oxysterol-binding protein homologue found in yeast that is essential for the intracellular transport of sterols and cell life. In this study, molecular dynamics simulations were used to investigate the binding of ergosterol, 25-hydroxycholesterol, and lipid moieties to Osh4. The binding energies between both sterols and Osh4 were dominated by van der Waals interactions with residues within the sterol binding pocket, and were further stabilized by water-mediated interactions with polar residues at the bottom of the binding pocket (W46, Q96, Y97, N165, Q181). Only Q96 was able to form direct hydrogen bonds with each sterol. Possible lipid binding sites on the surface of Osh4 were identified by docking four lipid moieties modeled from different lipid head groups (phosphatidylcholine, phosphatidylserine, phosphatidylinositol(4,5)biphosphate, and phosphatidylinositol(3,4,5)triphosphate). Though lipids docked to several regions along the protein surface, the most commonly encountered regions were found along the β-barrel region, a loop formed by residues 236-244, and the mouth of the sterol binding pocket. Several residues identified in these regions, such as K168, A169, K173, and E412, are either included or adjacent to residues experimentally implicated in Osh4 membrane binding. Lipid docking did not result in favorable binding regions on the R7 helix or the distal binding surface, suggesting that protein backbone conformational changes are needed for membrane attachment in these regions. Ultimately, understanding how Osh4 attaches to cellular membranes and binds sterol will lead to a clear understanding of how this protein transports sterols between organelles in ViVo. Introduction Sterols, such as cholesterol in mammalian cells and ergosterol in yeast, are known to affect both structural and signaling properties of cellular membranes.1 Because sterol concentration is not homogeneous within the cell, a concentration gradient must be tightly maintained across individual organelles. For example, ergosterol constitutes approximately one-third of the yeast plasma membrane (PM) by molar concentration yet less than one-tenth of the endoplasmic reticulum (ER).2 A highly precise and incompletely understood sterol sorting mechanism allows for this disparity in sterol concentration from the ER, where sterols are synthesized, to the PM.3 Sterols are transported, in part, by vesicular mechanisms through the Golgi.1,3,4 Upon movement to the Golgi, sterols associate with sphingolipids to form lipid rafts and are then transported to the PM via secretory vesicles along cytoskeleton tracks in an ATP dependent manner. Nonvesicular sterol transportation is also thought to occur by means of lipid transport proteins (LTPs), where soluble proteins bind to sterols and carry them from the ER to the PM.1,3 Lipid transfer between membranes through a protein carrier has already been identified for ceramide via the ceramide transfer protein (CERT).5 The oxysterol-binding protein (OSBP), related proteins (ORPs), as well as their yeast (Saccharomyces cereVisiae) homologues (OSHs) form a large family of LTPs that have been implicated in nonvesicular sterol transport between membranes.6 These proteins are characterized by the presence of an OSBPrelated domain (ORD), which is capable of binding oxysterol ligands.7 The yeast genome houses seven Osh genes, and yeast remains viable following the deletion of any one Osh protein, * To whom correspondence should be addressed. E-mail: jbklauda@ umd.edu. Phone: (301) 405-1320. Fax: (301) 314-9126.

Figure 1. The structure of the Osh4 protein divided by subdomain (25 ns snapshots). A, Osh4 complexed with ergosterol. B, Osh4 complexed with 25-hydroxycholesterol. The protein regions are colorcoded as follows: lid region (residues 1-29) - red; central helices region (30-116) - orange; β-barrel region (117-307) - green; C-terminal region (308-434) - cyan.

though its function is impaired. However, deletion of all seven proteins results in cell death and implies an overlapping function among all Osh proteins.8 Osh4 is, by far, the most highly expressed Osh protein in yeast with an abundance of approximately 32 000 molecules per cell.9 Currently, Osh4 is the only OSBP protein crystal structure that has been solved, with structures available for the protein complexed with cholesterol, ergosterol, and three hydroxycholesterols.10 Sterols bind inside of a tunnel formed by a 19-strand β-sheet that nearly forms a complete β-barrel (Figure 1). A flexible N-terminal lid domain occludes the bound sterol from the aqueous phase. Recently, it has been suggested that the N-terminal lid forms an ArfGAP1 lipid packing sensor (ALPS) motif, a membrane binding motif that preferentially targets membranes with a high positive curvature (38 versus 90 nm liposomes).11 This lid is thought to

10.1021/jp106890e  2010 American Chemical Society Published on Web 10/06/2010

Sterol Binding and Membrane Lipid Attachment bind to membranes in the sterol free (apo) or open state, allowing for sterol uptake from the membrane. While the ALPS motif is not found on most ORPs, several ORPs contain other functional domains located in the Nterminus with respect to the ORD. For example, the Osh1-3 proteins as well as human OSBP and several other human ORPs contain a Plecstrin homology (PH) domain.12 Many, though not all, PH domains can bind phosphoinositides (PIPs) with varying degrees of affinity.13 While Osh4 lacks a PH domain, its ORD is capable of binding to PIPs.7 Furthermore, the presence of phosphatidylinositol(4,5)biphosphate (PIP2) has been shown to stimulate cholesterol transfer between donor and acceptor liposomes in Vitro, and may possibly serve as a means for the regulation of sterol distribution between cellular compartments by ORPs.6 Though the mechanism of interaction between PIPs and Osh4 remains unclear, it is thought that PIP binding occurs on regions of the external surface of the protein. A triple glutamate (R236E/K242E/K243E) Osh4 variant is incapable of binding to PIPs, while alterations to charged residues near the mouth of the sterol-binding pocket do not affect PIPs’ ability to stimulate sterol transfer between membranes, suggesting that the flexible 236-244 surface loop may be important with regard to Osh4’s ability to attach to PIP membranes. However, due to the location of this loop away from the mouth of the Osh4 binding pocket, it is unclear how PIP binding to this region would help facilitate sterol extraction and delivery. Molecular dynamics (MD) simulation is a useful tool for investigating ligand binding with atomistic detail.14-16 This tool has been used more recently to investigate the dynamics of the Osh4 protein in Singh et al.17 and Canagarajah et al.18 In Singh et al., water-mediated interactions between the ring hydroxyl group of cholesterol and polar residues in the Osh4 binding pocket were found to be significant for sterol binding, while the lid had a negligible effect on stabilizing the bound sterol within the binding pocket. Additionally, a mechanism for sterol release and uptake from the cytoplasm was derived and conceptualized as a dual molecular ladder.17 Stepwise cholesterol unbinding was also observed in Canagarajah et al.18 where the rate limiting step in sterol exchange was identified as the lid opening event. In Canagarajah et al., the Osh4 R7 helix was suggested to exist in a mobile, metastable state, while the lid was closed and suggested to exist in a lower mobility, stable state in the apo conformation. This study aims to further investigate the energetics of sterol binding through MD simulations of the Osh4 protein complexed with yeast’s natural sterol (ergosterol) and a hydroxysterol (25hydroxycholesterol). Region specific backbone structural changes of the Osh4 protein were also examined over the course of these simulations in order to identify conformations not observable in the crystal structure. In addition, docking studies were used to probe the protein surface for regions that have affinity toward certain phospholipids commonly found in yeast membranes: phosphatidylcholine (PC), phosphatidylserine (PS), phosphatidylinositol(4,5)biphosphate (PIP2), and phosphatidylinositol(3,4,5)triphosphate (PIP3). Blind docking was used to search for sites on a protein that will bind to a specific ligand, and was effective in identifying binding sites for drug-sized ligands in proteins with up to 1000 amino acids.19 Blind docking was used to determine potential membrane binding sites on the Osh4 surface based on lipid headgroup moieties. Methods Molecular Dynamics. Five MD simulations were completed on the Osh4 protein complexed with ergosterol and five with

J. Phys. Chem. B, Vol. 114, No. 42, 2010 13563 25-hydroxycholesterol using the CHARMM20 and NAMD21 packages. The initial X-ray crystallographic structure of these protein complexes (1ZHZ for Osh4/ergosterol and 1ZHX for Osh4/25-hydroxycholesterol) as well as the residue numbering scheme is taken from Im et al.10 All crystallographic waters as well as liganded sterol residues are maintained. The third residue found in both crystal structures, P1, is referred to as the first residue even though the initial crystal structures taken from the PDB contain two additional residues (M-1 and D0). All visual representations of molecular systems are constructed using the visual molecular dynamics (VMD)22 program. Appropriate CHARMM patches were applied to the C-terminal and Nterminal residues. The CHARMM C22/CMAP23,24 force field was used for all simulations, and ergosterol and 25-hydroxycholesterol were parametrized by using existing cholesterol parameters25 and partial charges as a reference. Both structures were solvated using a pre-equilibrated water box consisting of TIP3P model water molecules26 in CHARMM, forming a 100 × 100 × 100 Å cubic unit cell. Minimization was conducted in CHARMM through the steepest descent (SD) algorithm followed by adopted basis Newton-Rhapson (ABNR) minimization in order to reduce unfavorable energy contacts. As the protein contains an initial charge of -10, the system was neutralized and solvated in CHARMM using a 0.15 M NaCl solution. The final system for the solvated, ergosterol-complexed Osh4 protein consisted of 1 protein, 1 ergosterol residue, 34 chlorine ions, 44 sodium ions, and 29 715 TIP3P water molecules for a total size of 96 419 atoms. The final system for the solvated, 25-hydroxycholesterol-complexed Osh4 protein consisted of 1 protein, 1 25-hydroxycholesterol residue, 34 chorine ions, 43 sodium ions, and 29 737 TIP3P water molecules for a total size of 96 339 atoms. Both systems were heated in CHARMM from 110.15 to 310.15 K over a period of 100 ps using a 1-fs integrator time step. The final temperature of 310.15 K was selected in order to mimic the experimental setup of Singh et al.17 Five production runs for each Osh4-sterol complex were conducted in NAMD with different initial velocity seeds for the purpose of data collection. For each simulation, the heated structure was allowed to thermally equilibrate for a period of 500 ps using a 2 fs integrator time step. During this equilibration period, the pressure was held constant at 1 bar using a Langevin piston and the temperature was rescaled every 500 time steps. Following equilibration, 25 ns of constant pressure, temperature, and molecular (NPT) simulation was completed on each system. Pressure was maintained at 1 bar using a Langevin piston, and the temperature was maintained at 310.15 K using Langevin dynamics. For all simulations, Lennard-Jones interactions were smoothed by a switching function over 10-12 Å and particle mesh Ewald (PME)27 was used to compute long-range electrostatics. Periodic boundary conditions were used for all simulations, and all hydrogen bond lengths were constrained using the SHAKE algorithm28 (CHARMM) or the RATTLE algorithm29 (NAMD). In total, these simulations constitute 0.25 µs of production run simulations. Docking and Lipid Moiety Simulations with Osh4. Blind Docking. Blind docking was used to investigate potential regions of the Osh4 protein that are favorable toward binding with PIP lipids as well as PC and PS lipids using AutoDock4 (AD4) with AutoDockTools4.30 PIP2 was selected as a test ligand because it has been demonstrated that Osh4 possesses the ability to bind to this specific PIP.7 Furthermore, PIP2 is believed to be the primary PIP lipid responsible for stimulating sterol transport between cellular membranes based on in Vitro experiments that

13564

J. Phys. Chem. B, Vol. 114, No. 42, 2010

demonstrated the presence of PIP2 increases the cholesterol transportation rate by Osh4 between donor and acceptor vesicles.6 Though yeast does not contain detectable levels of PIP3, the ligand was chosen as the second PIP test ligand based on its ability to also stimulate cholesterol transport in Vitro.6 No other PI/PIP lipid was able to stimulate cholesterol transport to any notable degree.6 PC was selected because it is enriched in the yeast ER, and PS was selected because it is enriched in the PM.2 Furthermore, increased PS membrane concentration has been correlated to an increase in sterol transfer between liposomes in Vitro.31 Model head groups were constructed from these lipids by truncating each lipid at the C2 carbon (linker to the two alipathic side chains). Coordinates for the PC and PS models were derived from 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC) and 1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-L-serine (POPS) molecules used in previous MD simulations of a model yeast membrane.32 PIP2 and PIP3 coordinates were similarly derived from their corresponding structures from Li et al.33 In order to sample multiple protein structural conformations as well as side chain positions, 5 snapshots were taken in 5 ns intervals from two Osh4/ergosterol MD simulations and two Osh4/25-hydroxycholesterol MD simulations as well as the crystal structure of Osh4 bound to ergosterol and that of Osh4 bound to 25hydroxycholesterol for a total of 22 conformational snapshots. All bound sterol ligands and solvent molecules were removed prior to docking. Additionally, each ligand was docked against the crystal structures of the engineered “lidless” Osh4 protein provided from Im et al.10 This structure (1ZI7) contains three distinct coordinate sets and exists in an unliganded state, with residues 1-29 deleted and the flexible surface loop 236-240 replaced by an ectopic dipeptide sequence. Gasteiger-Marsili34 charges were applied to each model ligand, and were adjusted to obtain the proper integral charge on each phosphate group. Protein atomic partial charges were conserved from the MD simulations used to produce the coordinates. Ligand bonds were allowed to be freely rotatable, but receptor bonds were rigid. All docking tests used the Lamarckian genetic algorithm35 for searching and the default AD4 parameter set. The surface of the protein was searched using a 120 × 120 × 120 point grid with a grid spacing of 0.625 Å. Each conformational snapshot was docked 25 times for 25 million iterations per instance for each ligand studied, producing a total of 550 results. Each engineered lidless conformation was docked 50 times, producing an additional 150 results. For each model compound tested, docking results for all conformations were pooled and the 20 conformations pertaining to the lowest AD4 free energy of binding (∆Gbind) were selected for further analysis. These conformations will be referred to as the “top results”. Because multiple receptor conformations were sampled for each model, clustering results by root-mean-square displacement (RMSD) is not an ideal way to detect different binding sites for these tests. Instead, potential binding regions are segregated on the basis of similar interacting residues. All results from native Osh4 structures were pooled separately from engineered lidless results. A detailed explanation on how AD4 calculates binding free energy can be found in Huey et al.36 Docking with Flexible Side Chains and MD with PIP2. Docking with select flexible side chains was conducted on two regions of the protein using the PIP2 model ligand. The first region was defined by an area on the Osh4 surface that showed the highest tendency to dock model lipids in the blind docking tests, and confined by a 100 × 100 × 100 point grid (grid

Rogaski et al. spacing of 0.375 Å) centered near K180. As AD4 has a limit of 32 freely rotatable bonds for any given system, only select basic residues (K168, K180, K407, K411) were set to have freely rotatable side chains. These residues had the highest tendency to interact with model ligands during the blind docking tests. The protein conformation most favorable toward PIP interaction during the blind docking tests was chosen as the receptor, and the PIP2 model was docked to this conformation 400 times with 2.5 million iterations per instance. Docked conformations were ranked by ∆Gbind, and the best 40 conformations were chosen as the “top results”. Two docked conformations from the top results were selected for MD testing to ensure stability of the docked conformation. PIP2 parameters and atomic partial charges are taken from Li et al.33 and were modified to account for truncation at the C2 carbon. The PIP2 model was subject to SD and ABNR minimization, while the protein atoms were fully constrained. Structures were solvated in a 100 × 100 × 100 Å TIP3P water box and equilibrated for 100 ps in NAMD at 310.15 K with temperature rescaling as described in the MD methods section. Structures were subject to a production run of 500 ps in NAMD in a manner identical to the solvated Osh4 protein with sterol ligand described in the Molecular Dynamics section. The second region analyzed with flexible residues was created by defining a sample space surrounding the 236-244 flexible surface loop with a 100 × 122 × 100 point grid (grid spacing of 0.375 Å) around the center of the loop. All appropriate side chains (R236, Y238, F239, S240, K242, K243, and N244) contained in the loop were set to contain freely rotatable bonds. Two protein conformational snapshots were used that pertained to different structural conformations of the loop (folded and extended). PIP2 was docked to each conformation 400 times with 2.5 million iterations per instance. Docked conformations were ranked by ∆Gbind, and the best 40 results were selected as the “top results”. This loop is thought to play an important role in Osh4’s ability to bind to PIPs.7 Results Molecular Dynamics. Protein Structure. The structure of the Osh4 protein did not change dramatically during the course of 25 ns for all simulations, as indicated by the RMSD of the Osh4 C-CR backbone atoms with respect to the X-ray crystallographic structures. Ergosterol and 25-hydroxycholesterol simulations were compared with the crystal structure obtained for Osh4 cocrystallized with ergosterol (1ZHZ) and 25hydroxycholesterol (1ZHX), respectively. Of the 10 production runs conducted, all RMSD values fell within the range 1.00-3.29 Å in ergosterol simulations and 0.98-3.09 Å in 25hydroxycholesterol simulations (Figure S1, Supporting Information). RMSD values were averaged during the final 10 ns of simulation, and were calculated to be 2.09 ( 0.26 and 2.00 ( 0.20 Å for simulations with ergosterol and 25-hydroxycholesterol, respectively. These RMSD values indicate that Osh4 is stable throughout the course of all simulations. In order to determine which regions of the protein contribute most to structural deviations from the X-ray crystal structure, the RMSD of backbone atoms in each subdomain was also investigated. Subdomains were divided in the same manner as in Im et al.10 and are displayed in Figure 1. For all simulations, the structure of the central helices and β-barrel subdomain do not deviate substantially from the X-ray crystal structure (Table S1, Supporting Information). The lid subdomain typically showed the widest range of RMSD values for all simulations (Figure S2, Supporting Information). Particularly high RMSD

Sterol Binding and Membrane Lipid Attachment

Figure 2. RMSD vs time for the MD1 simulation run of the Osh4 complexed with ergosterol. Subdomains are shown in Figure 1. These RMSD results were typical of all simulations studied (RMSD plots for other simulations not shown).

Figure 3. Conformational probability of the 367-381 loop. RMSD data is binned by 0.1 Å and is collected over the entire course of all 25 ns simulations. The total line (black) is the probability of RMSD for all five runs.

values were seen during one run (ergosterol MD1, Table S1, Supporting Information), where the lid RMSD drastically increased after approximately 12 ns, eventually reaching a plateau value of ∼6 Å during the last 5 ns of simulation (Figure 2). Though the orientation of the R1 helix (residues 9-21) remained stable during the course of these simulations, the first six residues of the N-terminus were found to be flexible and are responsible for the wide variation of RMSD values observed for the lid subdomain. The flexibility of these residues has also been observed in an MD study of the Osh4 protein complexed with cholesterol by Singh et al.17 The C-terminal subdomain also showed a wide variability in RMSD when compared with the central helices and β-barrel subdomains (Figure S2, Supporting Information). Conformational changes on a large surface loop consisting of residues 367-381 between the β18 sheet and R8 helix largely account for the observed variability in RMSD for this region. A range of RMSD values between 0.30 and 4.30 Å was calculated for ergosterol simulations, and a range of 0.34-4.02 Å was calculated for 25-hydroxycholesterol simulations. Binning of RMSD data indicates that a multitude of distinct structural conformations can exist in this loop, even at the short time scales presented by these studies (Figure 3). The RMSD of a surface loop consisting of residues 236-244 was also investigated, as this loop is thought to be potentially important with regard to membrane attachment.7,31 This loop was found in multiple conformations throughout the course of these simulations. Most conformational changes in this loop occurred on the order of picoseconds and were usually characterized by an RMSD change in the loop of ∼1 Å followed by a several nanosecond plateau at a new value, though several

J. Phys. Chem. B, Vol. 114, No. 42, 2010 13565 intermediate conformations existed that lasted on the order of tens to hundreds of picoseconds. Generally, two stable conformations for this loop were encountered: a folded conformation where the loop is folded upon itself and an extended conformation (Figure 4). The folded conformation was present in the crystal structure of both the ergosterol and 25-hydroxycholesterol complexes. Several hydrogen bonds stabilize this conformation, including S234-S240, S245-S240, and G235-G241 backbone hydrogen bonds. A backbone hydrogen bond between Y238 and G241 was also found in this conformation, though it was not as commonly encountered as the other bonds. A hydrogen bond between the R236 side chain and the G241 backbone was also encountered. In the extended conformation, most of the stabilizing hydrogen bonds found in the folded conformation are broken. The Y238-G241 backbone hydrogen bond remains present, and is more frequently encountered than in the folded conformation, and an S245-K242 backbone hydrogen bond is formed. Larger RMSD values for this region agree with the elevated mobility observed in MD simulations of Osh4 complexed with cholesterol in Canagarajah et al.18 and agree with higher than average B-factors found in this region of the crystal structure. Sterol Binding of Osh4. The binding energies (∆Ebind) of both ergosterol and 25-hydroxycholesterol were calculated using CHARMM routines for each simulation and then averaged. The ∆Ebind value of ergosterol was found to be -61.10 ( 1.26 kcal/ mol, while the ∆Ebind value of 25-hydroxycholesterol was found to be -63.52 ( 2.68 kcal/mol. For both sterols, van der Waals (VDW) interactions displayed a higher contribution to the total binding energy than electostatic interactions. For ergosterol, 83% of the total ∆Ebind stemmed from vdW interactions, while vdW interactions contributed to 76% of the observed ∆Ebind for 25hydroxycholesterol. In a manner consistent with Singh et al.,17 the average binding energy relative to the solvation energy (∆∆Ebind/sol) for each sterol was calculated. Solvation energies (∆Esol-wat) were found from a 10-ns MD simulation of each sterol explicitly solvated with TIP3P water and are shown in Table 1. Solvation energies in ethanol (∆Esol-eth) were similarly calculated and are shown in Table 1 along with the calculated ∆∆Ebind/sol. From experiments conducted in ethanol from Im et al.,10 25hydroxysterol possesses a modestly higher affinity toward Osh4 than cholesterol (a Kd of 55 nM for 25-hydroxycholesterol versus a Kd of 300 nM for cholesterol). The binding energy calculations presented here do not include entropic contributions to the free energy term but rather provide an estimation of the enthalpic term. Entropic differences between each sterol in solvent may play an important role in estimating ∆∆Gbind/sol-eth. However, these calculations do show a large difference between ∆∆Ebind/sol-wat and ∆∆Ebind/sol-eth for 25-hydroxycholesterol, suggesting that the difference in Osh4’s affinity for 25-hydroxycholesterol over ergosterol may be reduced in a water solvent. Residue specific sterol-protein interaction energies are shown in Table 2. Ergosterol/25-hydroxycholesterol data collected in this study are also compared to interaction energy data presented for cholesterol complexed to Osh4.17 Nonpolar residues in the Osh4 binding tunnel showed similar interaction energies when compared with those of cholesterol for most cases. However, some anomalies existed due to structural differences between the investigated sterols. Interactions between individual sterols and charged residues near the mouth of the binding pocket (E107 and K108) varied greatly. Backbone atoms in K108 and K109 form vdW interactions on one side of each bound sterol’s tail, while hydrophobic residues (L241, L177, and I203) form vdW interactions on the other side. The E107 side chain is positioned

13566

J. Phys. Chem. B, Vol. 114, No. 42, 2010

Rogaski et al.

Figure 4. Two stable conformations of the 236-244 surface loop. A, The backbone atoms of the loop are kinked so the tip runs parallel to the Osh4 surface (folded conformation). Backbone hydrogen bonds between S234-S240, S245-S240, and G235-G241 are shown. Y238-G241 backbone hydrogen bonds are found in this conformation but are not shown. B, The backbone atoms of the loop adopt a more extended conformation that is more perpendicular to the Osh4 surface (extended conformation). Backbone hydrogen bonds between S245-K242 and Y238-G241 are shown. All hydrogen atoms were included in each simulation, though some are not displayed here. C, RMSD vs time for the production run used for snapshots A and B.

TABLE 1: Solvation Energies of Sterols in Water (wat) and Ethanol (eth) and Average Osh4 Binding Energies Relative to the Solvation Energiesa ligand sol-wat

∆E ∆Esol-eth ∆∆Ebind/sol-wat ∆∆Ebind/sol-eth a

ergosterol

25-hydroxycholesterol

-45.34 ( 0.37 -47.87 ( 1.03 -15.75 ( 1.31 -13.23 ( 1.63

-61.82 ( 0.53 -57.53 ( 1.71 -1.70 ( 2.73 -5.99 ( 3.18

All energies are reported in kcal/mol.

near the five-membered sterol ring. As the positioning of ergosterol and 25-hydroxycholesterol’s tails did not change significantly during MD simulation, the marked differences in interaction energy for L24, E107, and K108 can be explained by the differences in structure and composition among each sterol’s tail. Of the polar residues situated at the bottom of the binding pocket, Q96 displayed the most favorable interaction energies for ergosterol and 25-hydroxycholesterol. Additionally, this residue displayed the most favorable electrostatic interaction energy with ergosterol and 25-hydroxycholesterol, i.e., -2.4 and -2.5 kcal/mol, respectively. For all MD simulations, the standard deviation of the interaction energy between Q96 and each sterol was also significantly higher than those reported for other residues. Interaction energy data for all simulations were combined and sorted by ligand, and were than binned by 0.1 kcal/mol intervals. Through binning, Q96 displayed two distinct energetic peaks separated by approximately 5 kcal/mol (Figure 5).

Binned data of Q96 sterol interaction was fitted to multiple Gaussian distributions in order to account for the probability associated with each dominant peak. By integrating all peaks associated with each energy state, the lower energy state was encountered 21% of the time for ergosterol simulations with the most probable energy being -9.2 kcal/mol, while the higher energy state was encountered 79% of the time and reached a maximum frequency at -4.1 kcal/mol. For the Osh4/25hydroxycholesterol simulations, the lower energy state was more populated, being encountered 29% of the time with the maximum frequency occurring at -9.1 kcal/mol, while the higher energy state was less favorable than with ergosterol, being most frequent at -3.4 kcal/mol and encountered 71% of the time (Figure 5). The lower energy peaks are attributed to direct hydrogen bonding between Q96 and the C3-OH group on each sterol, while the higher energy peaks correspond to nonhydrogen bonded configurations or water-mediated hydrogen bonded configurations (Figure 6). A similar analysis was conducted on all other Osh4 residues near the sterol’s C3-OH group, and no other residue exhibited the same two-peak pattern as Q96. However, the interaction energy of W46 displayed a highly skewed distribution (Figure 5) possibly containing a small peak obscured under the tail of the larger peak. Direct hydrogen bonding between both sterols’ ring OH group and W46 was observed on occasion, but these bonds were short in duration and did not achieve the same level of stability of the Q96 hydrogen bond. Docking and Lipid Moiety Simulations with Osh4. Blind Docking. Four model ligands (PC, PS, PIP2, and PIP3) were docked against several conformational snapshots of the Osh4 protein, and the regions found to be important are color-coded

Sterol Binding and Membrane Lipid Attachment

J. Phys. Chem. B, Vol. 114, No. 42, 2010 13567

TABLE 2: Averaged Interaction Energies between Sterols and Nearby Residues from MD Simulationa nonpolar residues

ergosterol

25-hydroxycholesterol

cholesterol

F13 L24 L27 I33 L39 F42 P110 I167 L177 V179 L201 I203 I206 P211 V213

-0.7 ( 0.1 -1.2 ( 0.2 -0.4 ( 0.2 -1.2 ( 0.1 -1.9 ( 0.1 -4.0 ( 0.2 -1.9 ( 0.1 -1.8 ( 0.1 -1.4 ( 0.3 -1.1 ( 0.2 -1.4 ( 0.1 -2.3 ( 0.2 -1.1 ( 0.2 -1.0 ( 0.1 -0.8 ( 0.2

-0.9 ( 0.1 -1.5 ( 0.4 -0.6 ( 0.3 -1.0 ( 0.2 -2.0 ( 0.2 -3.7 ( 0.3 -2.4 ( 0.1 -2.1 ( 0.2 -1.6 ( 0.3 -1.2 ( 0.2 -1.7 ( 0.2 -2.5 ( 0.4 -1.2 ( 0.3 -0.9 ( 0.2 -1.0 ( 0.1

-0.9 ( 0.7 -2.2 ( 0.4 -0.7 ( 0.2 -1.0 ( 0.1 -2.0 ( 0.2 -4.1 ( 0.3 -2.2 ( 0.2 -2.4 ( 0.1 -1.1 ( 0.2 -1.0 ( 0.2 -1.6 ( 0.2 -2.6 ( 0.3 -1.0 ( 0.2 -0.9 ( 0.1 -1.1 ( 0.1

polar residues

ergosterol

25-hydroxycholesterol

cholesterol

W46 Q96 Y97 N165 Q181

-1.0 ( 0.5 -4.9 ( 1.3 -3.1 ( 0.2 -2.1 ( 0.3 -2.1 ( 0.3

-1.2 ( 0.7 -4.8 ( 1.6 -3.7 ( 0.3 -2.0 ( 0.3 -2.0 ( 0.5

-1.1 ( 0.5 -4.5 ( 0.9 -3.9 ( 0.3 -2.1 ( 0.5 -2.7 ( 0.5

charged residues

ergosterol

25-hydroxycholesterol

cholesterol

E107 K108 K109

-4.0 ( 0.2 -2.4 ( 0.2 -2.7 ( 0.1

-3.3 ( 0.7 -2.1 ( 0.3 -2.4 ( 0.4

-5.1 ( 0.2 -1.1 ( 0.2 -2.7 ( 0.2

a Only residues with interaction energies above approximately 1 kcal/mol are reported. Most of the cholesterol data is taken from Singh et al.,17 where electrostatic interactions are not included in the interaction energies for nonpolar residues in the cholesterol data set. However, those in italics are calculated for this study on the basis of previous trajectories and electrostatic interaction is included for the nonpolar residues.

Figure 5. The probability distributions of hydrogen bonding energetics for Q96 and W46.

in Figure S3 (Supporting Information). Of the model lipid compounds tested, the PC model presented the greatest diversity in terms of total number of binding sites identified in the top selected results. For this model, four different binding regions (Figure 7) were identified when docked against the crystal and MD snapshot conformations (Table 3). Two of these regions were also identified during the PIP2 tests, while only one of

these regions was found during the PIP3 and PS tests (Table 3). No regions were identified during the PS, PIP2, or PIP3 tests that were not also identified during PC blind docking. High negative charges associated with the PIP models (-4 for PIP2 and -6 for PIP3) prevented docking to PC sites where near equal distributions of positively and negatively charged amino acids were encountered. This reduced the number of docked regions identified in PIP models compared to the PC model. However, given the electroneutrality of the PC model and its small size, nearly 50% of the results had to be rejected due to docking inside of the sterol binding pocket of the Osh4 protein. The docking of model lipids to within the sterol binding pocket occurred as a result of a vacant sterol binding pocket (see Methods). This issue was much more prevalent in conformations taken from MD snapshots of Osh4 complexed with 25hydroxycholesterol than in conformations taken from Osh4 complexed with ergosterol. There were no instances of either of the PIP models docking inside of the Osh4 sterol binding pocket during these tests, and instances where PS docked inside of the pocket were rare. The four binding regions identified for the PC model are color-coded in Figures 7 and S3 (Supporting Information). Among all of the top results, the R3 region produced the conformation exhibiting the most favorable ∆Gbind (-3.26 kcal/ mol). However, only 2 out of the 20 selected results for this test were contained in this region. The R3 region is located near the distal side of the protein, interacting with residues on the solvent exposed portion of the R3 and R4 helices, as well as the surface loop that connects them. Both conformations in this region were stabilized through electrostatic interactions between a glutamic acid residue (E51/E59) and the choline group of PC as well as electrostatic interactions between a lysine side chain (K87) and the phosphate group of PC (Figure S4A, Supporting Information). Another PC binding site (C-terminus region) was stabilized through electrostatic interactions with residue types similar to those found in the R3 region. This region was not frequently encountered during testing, constituting only one of the top results with a ∆Gbind value of -2.49 kcal/mol (Figure S4B, Supporting Information). Similarly, the R4-R6 region only constituted one of the top results with a ∆Gbind value of -2.46 kcal/mol. Of these three sites, only the R4-R6 region was also encountered when docking with PIP models. This region appeared during PIP2 tests but was only found in one of the top results for this model (-4.08 kcal/mol, Figure S4C, Supporting Information) and did not appear in any of the PIP3 top results. The β-crease region was located near the mouth of the sterol binding pocket, centered near K180, and was defined by a crease between a solvent accessible portion of the β-barrel and a large surface loop near the C-terminal end of the protein. This site was the most prevalent binding site encountered during blind docking tests of the native Osh4 protein surface for all model ligands, being encountered in 16 top results for PC (-2.88 kcal/ mol, most favorable conformation), 19 cases for PIP2 (-4.86 kcal/mol, most favorable), and all 20 cases for PIP3 (-5.30 kcal/ mol, most favorable) and PS (-5.22 kcal/mol, most favorable). Though nearly all of the lowest energy PIP2 and PIP3 conformations docked within this region, several receptor conformations taken from MD snapshots failed to dock PIP2 or PIP3 in this region favorably. For these snapshots, PIP2 and PIP3 favored the R4-R6 region. However, because ∆Gbind was typically unfavorable for the R4-R6 region in most cases, the region is lowly populated in the top results for PIP2 (1 result out of 20) and unpopulated for PIP3. The RMSD analysis presented in the Protein Structure section demonstrates that the β-barrel, which

13568

J. Phys. Chem. B, Vol. 114, No. 42, 2010

Rogaski et al.

Figure 6. Sample binding conformations of ergosterol and 25-hydroxycholesterol with Osh4. A, 25-hydroxycholesterol forming a direct hydrogen bond with W46 as well as a water-mediated hydrogen bond with Q181. B, 25-hyroxycholesterol forming a direct hydrogen bond with Q96 as well as a water-mediated hydrogen with Q181. C, Ergosterol forming a water-mediated hydrogen bond with Q96. D, Ergosterol forming multiple watermediated hydrogen bonds with W46, Q96, and Q181. Nonpolar hydrogens were simulated but are not shown. Hydrogen bonds are shown as dashed lines.

partially defines the β-crease region, does not structurally deviate significantly with respect to the protein’s crystal structure. A further RMSD analysis on the second section partially defining the β-crease region, the 393-416 loop, also yielded no significant deviations from the protein’s crystal structure (RMSD