Analysis of the Structural and Dynamic Properties of Human N

8730

J. Phys. Chem. B 2008, 112, 8730–8736

Analysis of the Structural and Dynamic Properties of Human N-Terminal Domain of Apolipoprotein E by Molecular Dynamics Simulations Isabelle Ortmans† and Martine Pre´vost* Structure et Fonction des Membranes Biologiques, UniVersite´ Libre de Bruxelles, CP 206/2, Bld du Triomphe, B-1050 Bruxelles, Belgium ReceiVed: January 11, 2008; ReVised Manuscript ReceiVed: March 25, 2008

Whereas the lipid-free N-terminal domain of apolipoprotein E (apoE-NT) adopts a four-helix bundle, the lipid-bound form is believed to undergo a large conformational change likely to be characterized by the opening of the bundle. ApoE-NT in a water/alcohol mixture was also shown to experience conformational changes exhibiting similarities with those induced upon lipid binding. The structure and dynamics of apoENT have been here investigated by analyzing 40 ns and 60 ns molecular dynamics simulations performed in water and in a water/propanol mixture, respectively. The overall structural properties show alterations of the tertiary structure of apoE-NT in the water/alcohol system in agreement with those observed experimentally. In contrast, in the water simulation, the sampled conformations remain closer to the crystal structure that served as a starting point for both simulations. Interestingly, several propanol molecules are seen to penetrate two hydrophobic regions of the bundle interior. One of these regions is enclosed in part by the short helix (H1′) connecting helices 1 and 2 of the bundle which has been experimentally shown to be important for modulating lipid binding activity of apoE-NT. Principal component analysis of the water/propanol trajectory confirms that the region including H1′ is the locus of the largest motion. Another region involves the loop connecting helix 2 and helix 3 which has been hypothesized to play the role of a hinge in the opening of the bundle. Introduction Human apolipoprotein E (apoE), which stabilizes lipoprotein particles, plays an important role in regulating cholesterol homeostasis as well as in regulating plasma triglyceride clearance through its interaction with members of the lipoprotein receptor family such as the low-density lipoprotein receptor (LDLR).1,2 As a result of its widespread presence in amyloid plaques, it has been suggested that apoE may also function as a pathological chaperone protein, that is, one that induces β-pleated conformation in amyloidogenic polypeptides.3 ApoE is composed of two structurally and functionally independent domains.4,5 The N-terminal domain (apoE-NT) contains the LDLR-binding region.2 In absence of lipids, apoE does not recognize LDLR while apoE-NT domain bound to lipids binds efficiently to the receptor.6 This suggests that a lipid-bindinginduced conformational change in the N-terminal domain is essential for apoE to serve as a ligand for LDLR. The crystal lipid-free form of the apoE-NT7 consists of a four-helix bundle in which the nonpolar faces of the helices are directed toward the center of the bundle (see Figure 1). The interaction of apoENT with lipids is believed to occur by the opening of the fourhelix bundle exposing the hydrophobic residues thereby favoring the binding to the acyl chains of the lipid molecules.2,8,9 In that respect, an important breakthrough has been recently made by the determination, though at a low resolution (∼10 Å), of the lipid-bound form of apoE4 that features an extended conformation.10 * To whom correspondence should be addressed. Phone: 32 2 6502049; fax: 32 2 6505382; e-mail: [email protected]. † Present address: UCB-Pharma, Chemin du Foriest 1, B-1420 Brainel’Alleud, Belgium.

Figure 1. Ribbon diagram depicting the crystal structure of the apoENT domain. The four helices of the bundle are H1 (24-42), H2 (54-81), H3 (87-125), H4 (130-164). H1′ is the small helix connecting H1 and H2.

Investigation of the conformational changes in apoE upon binding to lipids is complex because of the size and heterogeneity of the lipoprotein particles in which apoE is found. In vitro apoE also interacts with lipids contained in discs or spherical

10.1021/jp8002678 CCC: $40.75  2008 American Chemical Society Published on Web 06/26/2008

Human N-Terminal Domain of Apolipoprotein E particles.11–13 These lipid systems though simpler than lipoprotein particles contain at least 2 protein molecules and 100 lipids.11 Recently, it was shown that trifluoroethanol (TFE)14 and a water-propanol mixture15 could induce structural changes, which present similarities with those observed in the presence of lipids. These observations suggest that these systems may be useful model systems for studying conformational changes occurring in apoE upon lipid binding. Because the function of apoE is intimately related to its structure and dynamics, it is important to understand the structural properties of this protein. Even though there is a large body of structural information on apoE-NT, characterization of the molecular factors likely to promote the bundle opening is difficult as this process is dynamic by nature. Molecular dynamics (MD) simulations provide a way in which to explore conformational dynamics of proteins with an extreme level of detail on the atomic motions. Now, with the advent of increasingly powerful computers and of improved simulation methodologies, trajectories of several dozens of nanoseconds are within reach. Although these time scales are still orders of magnitude smaller than those related to large conformational changes, MD simulations are expected to give a better understanding about the nascent steps which initiate the opening of the bundle upon lipid binding. Furthermore, a possible solution to the problem of sampling the conformational space, investigating the physical nature of protein dynamics, and finding functionally relevant motions is the use of collective coordinates which can be identified by principal component analysis (PCA). Collective motions within proteins can be derived from an analysis of many conformations, as can be generated by MD simulations, by separating them from small, presumably uninteresting motions. Recent studies have shown that functionally relevant motions occur along the direction of a few collective coordinates,whichdominantlycontributetoatomicfluctuations.16–18 One example is the transition from an open to a closed conformation found during an MD simulation of the cAMP receptor protein that occurs upon binding to DNA.19 Interdomain dynamics of a drug transport accessory protein has also been studied using principal component analysis.20 The structure of apoE-NT has been previously studied by MD simulations in an aqueous solution and at a water-organic phase interface.21,22 The objective of the study here is to examine the conformational dynamics of apoE-NT in two different conditions mimicking either a lipid-free or a lipid-bound environment of the protein. We evaluated the effects of adding n-propanol to an aqueous solution on several structural and dynamic properties of apoE-NT and compared them to its properties in water. To this end, two MD simulations of duration of 60 ns in a water/propanol mixture and of 40 ns in water were performed to analyze the motions of apoE-NT. PCA was carried out to examine the long-time scale collective motions. Methods Simulation Protocols. Energy minimization, equilibration, and production dynamics runs for the two systems were performed using the CHARMM program23 starting with the high-resolution crystallographic four-helix bundle structure of apoE-NT7 in which residue portions 1-22 and 165-191 are not seen in the electron density of the X-ray experiments indicating that these two regions are disordered. The protein and water interacted via the CHARMM 22 force field where all protein atoms are explicitly represented24 and where the water is portrayed by the TIP3P model.25 The TIP3P water model has the great advantage that it forms a consistent

J. Phys. Chem. B, Vol. 112, No. 29, 2008 8731 optimized set with the protein parameters in the CHARMM force field. Bonds connecting hydrogens were constrained using the SHAKE algorithm26 which permitted the use of an integration time step of 1 fs. For van der Waals terms, standard nonbonded truncation was employed with an atom-based 12 Å cutoff distance and a switching function being applied between 11 and 12 Å. To avoid truncation of long-range electrostatic interactions, the particle-mesh Ewald method was used. First, for each system, the protein was minimized for 100 steps using the steepest descent algorithm to relieve bad crystal contacts. The simulation in water was carried out on a system that consisted of the protein plus 121 crystal water molecules and 4570 built water molecules. After 40 ps equilibration of the water molecules in the presence of the fixed protein and crystal water molecules, the system was slowly brought up to a temperature of 300 K for 300 ps and was equilibrated for 500 ps at constant pressure and temperature (P ) 1 atm). Ten nanoseconds of equilibration was then carried out in the canonical ensemble using the periodic volume of the previously equilibrated system (dimensions ) 84.45/45.09/42.16 Å). A subsequent run of 40 ns was then produced for analysis. The simulated water/40% propanol mixture comprised the protein and crystal water molecules solvated by 2189 additional water and 370 propanol molecules. The bulk water was first equilibrated for 125 ps with the protein and propanol kept fixed. The propanol was then relaxed together with the bulk water for 425 ps with the protein being fixed. The whole system was then heated up and equilibrated for 125 ps at constant pressure and temperature (P ) 1 atm). Using the volume reached by the system at the end of this simulation (dimensions ) 80.82/44.83/ 38.09 Å), 10 ns of equilibration was carried out in the canonical ensemble. A subsequent run of 60 ns was then produced for analysis. Simulation Analysis. PCA is a tool for identifying largescale collective fluctuations from MD trajectories or from any set of structures. The method is based on the calculation and diagonalization of the covariance matrix of atomic fluctuations which yields a set of eigenvalues and eigenvectors. The eigenvectors of this matrix represent a set of directions in the 3N-dimensional space along which the fluctuations observed in the simulation are uncoupled with respect to each other. The eigenvalues quantify the fluctuations along the respective eigenvectors. The motions along the principal modes can be further analyzed by projecting the atomic trajectories onto the corresponding eigenvectors. Analysis was performed on the two trajectories using Gromacs program.27 Results We evaluated the effects of adding n-propanol to an aqueous solution on several structural and dynamic properties of apoENT. Overall Structural Properties. As a first check of the stability of the simulations and to assess the extent of the motions quantitatively, the root-mean-square deviation (rmsd) from the starting conformation of the CR positions of the protein was computed along the 60 ns and 40 ns trajectories (see Figure 2). The starting rmsd value is similar in the water (apowat) and in the water/propanol mixture (apoprop) trajectories with a value of about 2 Å (which corresponds to a system equilibrated for 10 ns; see Methods). In apoprop, there is a steady increase up to 15 ns where the rmsd reaches a value of 2.7 Å, while in apowat, the rmsd decreases to 1.5 Å for 10 ns. At 40 ns, the rmsd is about 3 Å in apoprop and 2 Å in apowat. The rmsd in apoprop then fluctuates for the last 20 ns around a value of 3 Å.

8732 J. Phys. Chem. B, Vol. 112, No. 29, 2008

Figure 2. The rmsd of the CR atoms of the bundle helices computed from the crystal structure as a function of the simulation time of the apoprop (black) and apowat (green) trajectories.

Figure 3. Solvent-accessible surface area of the side chains (SASA) averaged over the last 20 ns of apoprop (blue) and apowat (red) trajectories. Locations of the helices are indicated by rectangles.

Secondary Structures. The time evolution of the secondary structures was examined. The content in R-helical structure is maintained along the apoprop and apowat trajectories (data not shown). This correlates with Fourier transform infrared (FTIR)9 and circular dichroism (CD)4 measurements that indicate that apoE (1-183) shows no real change in the secondary structure content upon binding to lipid. It was, however, shown that the amount of R-helix in apoE (1-191) increases in the presence of n-propanol.15 Though this contrasted result could be ascribed to the different environments, it could also arise from the fact that the protein employed in the propanol studies is longer (1-191) than that used in these simulation studies (residues 23-166). The additional segments are indeed thought to be disorganized in aqueous solution and were shown to adopt a helical structure in a 126-183 protein fragment upon addition of TFE and dodecylphosphocholine (DPC).14,28 Accessible Surface Area of Side Chains. The solventaccessible surface area (SASA) of the side chains was computed and averaged over the last 20 ns for the two trajectories (see Figure 3). About 10% of the side chains in apoprop exhibit a significantly higher SASA relative to apowat. A few of these residues are located at the C-terminus of H1 and at the N-terminus of H2. Also, residues belonging to the C-terminus of H3 and to the N-terminus of H4 display a higher exposure to solvent. The six-residue loop connecting H2 to H3 exhibits a higher solvent accessibility in apoprop. This could be due to the hydrophobic content of this loop (Leu82, Pro84, Val85,

Ortmans and Pre´vost

Figure 4. Top: Number of contacts between the hydrophobic moiety of propanol molecules and the protein hydrophobic side chains averaged along the 60 ns apoprop trajectory. Bottom: Fraction of H bonds formed by side chain groups of apoE-NT with propanol molecules in apoprop trajectory. The fraction is defined as the number of conformations in which the corresponding group makes a hydrogen bond divided by the total number of conformations in the 60 ns trajectory. Locations of the helices are indicated by rectangles.

Ala86). Residues with the largest change in SASA in apoprop relative to apowat are W39, Q41, Q55, T83, T89, R92, S94, Q117, L126, L133, and K146. Protein-Propanol Interactions. The H bonds formed between the protein side chains and the propanol molecules are depicted in Figure 4. Hydrogen bonds are formed with almost all polar side chains, though for a few of them moderately, indicating that propanol solvates the protein surface in a rather homogeneous manner. The most hydrogen-bonded residue types are glutamine and glutamate each of which makes up about 10% of the total composition in aminoacids in apoE-NT. Glutamine residues, in particular, fulfill a dual role: aliphatic portions of their side chains allow binding to hydrophobic moieties whereas their polar tails can form hydrogen bonds with polar groups. Contacts between the hydrophobic moiety of propanol molecules and hydrophobic groups of protein side chains have also been monitored along the apoprop trajectory (Figure 4). Those contacts are spread along the whole protein sequence with some heterogeneity. The residues featuring the larger number of contacts with the lipophilic portion of propanol molecules are quite expectedly hydrophobic residues: L28, W34, W39, L52, A62, L63, L82, L93, A102, M125, L126, V135, L149, A160, V161, and A164. Polar or charged residues with a significant hydrophobic side chain portion are also involved in the formation of hydrophobic contacts with propanol: R32, Q58, K72, Y74, Q81, R92, K95, Q98, R103, Q117, Q123, R142, K146, Q156, K157, R158, and Q163. There is a larger number of hydrophobic contacts in two regions: one is enclosed by the C-terminus of H1 and the N-terminus of H2 and H1′, and the other region is located roughly at the other extremity of the bundle and is defined by the C-termini of H2 and H4 and the N-terminus of H3. The search for hydrogen bonds and hydrophobic contacts between protein atoms and propanol molecules revealed that a few propanol molecules penetrate the protein interior during the simulation. Two internal protein regions have been identified which are the main loci of propanol molecule dwelling: one region is located at the encounter of the C-terminal extremities of H1 and H3 and the N-termini of H2 and H4 and involves

Human N-Terminal Domain of Apolipoprotein E

Figure 5. View of one of the extremities of the bundle formed by the C-termini of H1 and H3 and the N-termini of H2 and H4 and by H1′ featuring one buried propanol molecule. Three structures are superimposed at different simulation times to show the rotation of Trp39 side chain which opens and closes the protein cavity and the displacement of H1′ which permits the entrance and exit of the propanol molecules. The bundle helices are depicted as cylinders. The protein side chains enclosing the buried cavity are shown as sticks and the propanol molecule is depicted as ball-and-sticks.

H1′. The interior of this strongly hydrophobic region is occupied for the last 20 ns successively by three different propanol molecules. Before entrance is observed, the interior is inaccessible from the solvent. In a first step, the side chain of Trp 39 rotates so as to facilitate the entrance of a propanol molecule (see Figure 5). The propanol can then go deeper into the protein interior to interact with buried hydrophobic residues causing a displacement of H1′ and an expansion of the cavity (see Figure 5). This is accompanied by a transient rotation of Trp39 side chain back to its original position occluding the protein cavity. After the Trp side chain rotation occurs, the propanol molecule to be buried is stabilized by hydrophobic contacts with hydrophobic portions of exposed residues and with other propanol molecules and by hydrogen bonds through its OH group mainly with surface water molecules and with the side chains of Trp39 and Glu50. This propanol molecule then penetrates the protein interior: the most buried hydrophobic residues (Leu51, Val56, Leu115, Leu137, and Leu141) interact with its hydrophobic portion while the OH propanol group forms a hydrogen bond with the buried carbonyl group of Tyr36. The exit of the propanol molecule and the entrance of another occur via the same mechanism and pattern of interactions. The other region accommodating propanol molecules is located at the interface between H2 and H3 near the loop connecting the two helices. One of the propanol molecules was observed to reside for 20 ns. The region can be depicted as a marked invagination that is lined on the surface by polar residues (Figure 6). Lys72, Gln98, and Gln101 help the entrance and exit of the propanol molecule by playing a dual role in forming a hydrogen bond with the OH group of the propanol and by making hydrophobic contacts as well with its lipophilic moiety. Like in the other region, other propanol and water molecules at the protein surface also play a role in interacting with the propanol molecule which gets into the protein interior. Large-Scale Motions. A useful approach in the identification of important motions of a protein is the use of PCA. We investigated the most significant collective modes of motion

J. Phys. Chem. B, Vol. 112, No. 29, 2008 8733

Figure 6. View of the deep invagination between H2 and H3 in which entrance and exit of propanol molecules is observed during apoprop simulation. Left: Molecular surface representation of one snapshot of apoE-NT in the apoprop simulation featuring one propanol molecule entering the open cavity. Right: Cartoon representation: The protein side chains enclosing the cavity are shown as sticks (color code: lys ) blue, gln and ser ) pink, leu and val ) green), and the propanol molecule is depicted as ball-and-sticks.

occurring during the apoprop and apowat simulations so as to decompose the global motion of the protein into a few principal motions. To compare the essential dynamics (ED) of two simulations of two similar systems, a method was developed which concatenates several trajectories fitted onto a same reference structure from which a covariance matrix is constructed.29 This method is a tool for evaluating similarities and differences between the essential motions in different trajectories of the same protein. Upon diagonalization of the covariance matrix of the combined trajectories, only a few vectors with the highest eigenvalues are obtained. The separate trajectories are then projected onto the eigenvectors from the combined trajectory featuring the highest eigenvalues and thus the largest fluctuations. The similarity between the two trajectories and the overlap of the sampled phase space can then be analyzed through the projections of the two simulations onto the plane defined by the first two eigenvectors (see Figure 7). First, there is an evident clustering into two groups of the two simulations. It can be seen that there is no overlap between apoprop and apowat simulations. Within apoprop trajectory, the sampled conformations explore two regions suggesting that there is a transition occurring during the apoprop simulation. The smaller region is sampled during the first 14 ns, and the second is sampled for the remaining 46 ns. This division occurs along both eigenvectors 1 and 2. Information can be obtained between differences in equilibrium structures and fluctuations of the separate trajectories along the eigenvectors obtained from the combined trajectory. Two types of properties have been analyzed. The structural and dynamic effects of the water/propanol mixture have been studied by considering the projections of the separate apoprop and apowat trajectories onto the eigenvectors calculated from the combined trajectories.29 The average projections and the mean

8734 J. Phys. Chem. B, Vol. 112, No. 29, 2008

Figure 7. 2D projection on the first two eigenvectors from the two simulations: apoprop and apowat. The distinct regions of the projections correspond to apowat (green) and apoprop (black from 0 to 14 ns and red from 14 to 60 ns).

Figure 8. (top) Average projection of the simulation time frames of the separate apoprop (red) and apowat (black) trajectories onto the eigenvectors calculated from the combined trajectories. (bottom) Mean square fluctuations in the projection of the simulation frames of the separate apoprop and apowat trajectories onto the eigenvectors calculated from the combined trajectories.

square fluctuations in these projections are shown in Figure 8. The main structural differences between apoE-NT in water and in the water/propanol mixture appear at eigenvector 1. On average, the apoprop and apowat simulations are at different positions along eigenvector 1 pointing out that there is a conformational change caused by the alcohol mixture. The structures extracted from the simulations corresponding to the minimum and maximum projections reveal one region which is the locus of a large conformational change located at the C-terminus of H1, the N-terminus of H2 and H1′, and the loop connecting H3-H4. H1 and H3 move more than the other two helices along the first eigenvector (Figure 9). The fluctuations in the projections show that in addition to the structural effect there are possible differences in dynamics along the directions described by eigenvector 1 and more markedly along eigenvector 2. The motion described by the latter (Figure 10) is larger in the loops connecting H2-H3 and H3-H4 and along H2 and H4 relative to H1 and H3. Visualization of the displacements along eigenvectors 1 and 2 using Dynatraj30 generates a porcupine plot (Figures 9 and 10). It can be seen that the first principal motion (Figure 9)

Ortmans and Pre´vost

Figure 9. View of the motions corresponding to the first eigenvector defined from the combined trajectories. (left) Superimposition of the minimum and maximum projection structures with eight interpolated frames. (right) Porcupine plot. Each CR has a cone attached pointing in the direction of motion described by the eigenvector for this atom. The two views are rotated from each other by 90° about the bundle axis.

Figure 10. View of the motions corresponding to the second eigenvector defined from the combined trajectories. (left) Superimposition of 10 structures describing the concerted motion. (right) Porcupine plot. Each CR has a cone attached pointing in the direction of motion described by the eigenvector for this atom. The two views are rotated from each other by 180° about the bundle axis.

corresponds to a sort of breathing motion between H1-H2 and H3-H4 accompanied by a large swing motion of H1′ and to a lesser extent by a sway of the loop connecting H3 and H4, which move away from the bundle. The second principal motion (Figure 10) can be depicted as a shear motion involving the sliding of H1-H2 pair onto H3-H4 pair along the helix bundle axis accompanied by a large swing motion of the loop connecting H2 and H3 and of the loop connecting H3 and H4. The rmsd was computed between the two structures corresponding to the maximum and minimum projections along eigenvector 1 by moving a window of a five-residue-long protein fragment (Figure 11). The two most structurally distant structures permit the identification of potential hinge residues which are important for generating the motions.31 There are a few regions that feature higher rmsd values. Peaks occur at the last turn of H1, at the connection between H1 and H1′ and the first turn of H1′, at the two residues connecting H1′ and H2 as well as the last three turns

Human N-Terminal Domain of Apolipoprotein E

Figure 11. rmsd calculated by moving a window of five sequential residues between the two structures corresponding to the maximum and minimum projection structures along eigenvector 1 (which are conformationally the most distant along that eigenvector; see Figure 9). Locations of the helices are indicated by rectangles.

in H3, and at the loop connecting H3-H4 and the first turn in H4. Those residues define a region including H1′ and its surrounding residues at one of the extremities of the bundle. This picture is in agreement with the motion along eigenvector 1 in which H1′ undergoes a large hinge motion around the C-terminus of H1 and the N-terminus of H2 together with the facing H3-H4 connecting loop. Other residues include the last turn in H2, the loop connecting H2 and H3, and the first two turns of H3. They delimit a zone which corresponds to the extremity of the bundle opposite to H1′. There is a third region of high rmsd values defined by residues in H3 (103-107) and H4 (145-150) that face each other. This corresponds to a bending of the H3-H4 pair at about the middle of the two helices. Discussion Our simulation results show that a significant alteration of the global structure as revealed by the rmsd occurs in the water/ propanol mixture simulation. These structural changes are in agreement with experimental data which show that the addition of n-propanol alters the tertiary structure of apoE-NT.15 In contrast, the protein structure in the water simulation remains close to the crystal structure. No extension of R-helical structure was observed in apoprop in agreement with CD and FTIR measurements of apoE-NT in the presence of lipids.4,9 An increase in R-helicity has, however, been experimentally observed in the presence of propanol. The experiments, however, were performed on the 1-191 fragment of apoE longer than our 23-166 fragment and which includes portions 1-22 and 167-191 known to be unstructured in water. The increase in secondary content experimentally observed in n-propanol is most likely to occur in these portions as suggested by the acquisition of helical structure observed in the 126-183 peptide when adding TFE or DPC to an aqueous solution.14,28 We monitored the propanol-protein interactions in apoprop to disclose a potential mechanism for alcohol-protein association which could be responsible for the alterations detected in apoprop simulation and for the larger conformational changes observed experimentally. To gain alcohol-protein interaction, propanol molecules have to get close to the protein whose surface is rather polar and charged and thus strongly hydrated. The mechanism by which the propanol molecules approach the protein surface occurs here through hydrogen bond interactions with the numerous charged and polar residues and the surface

J. Phys. Chem. B, Vol. 112, No. 29, 2008 8735 water molecules. The hydrophobic portion of these residues also interacts with the hydrophobic part of the propanol molecules. Indeed, the side chains of those polar and charged residues (glutamine, lysine, etc.) own a long aliphatic moiety. These typically nonlocal interactions (solvent-mediated and hydrophobic interactions) which have been suggested to be part of the mechanism of short-chain alcohol denaturation can lead to alteration of the tertiary structure.32 This mechanism by which propanol initiates and forms interactions with the helical protein in the apoprop simulation not only matches the association process proposed from numerous experiments to explain alcohol denaturation but also strikingly resembles the association mechanism revealed in MD simulations of a helical peptide in a zwitterionic micelle.33 Propanol molecules have also been shown to penetrate the fourhelix bundle interior. Two main internal regions are the locus of propanol molecule penetration. One is located at one extremity of the bundle and is enclosed by the C-termini of H1 and H3 and the N-termini of H2, H4, and by H1′. The other is located between H2 and H3 more toward the other extremity of the bundle. The entrance of propanol molecules has been monitored and is consistent with the data obtained from the hydrogen bond and hydrophobic contact analysis (see Results section). Propanol molecules at the protein surface hydrogen bond mainly to surface water and to charged or polar residues. Their hydrophobic portion is stabilized by interactions with other surrounding propanol molecules. As the propanol molecule moves into the protein interior, it interacts with the hydrophobic part of polar side chains. The propanol molecule then gets deeper into the hydrophobic core of the protein; interactions with the most buried aliphatic residues (Leu, Val) occur and the propanol hydroxyl group hydrogen bonds to buried polar groups, that is, backbone carbonyl oxygen. Exchanges of several propanol molecules within these two sites are observed and happen with a similar mechanism. The site enclosed by the C-termini of H1 and H3 and by the N-termini of H2, H4, and by H1′, which accommodates a buried propanol molecule, is a strong hydrophobic region of the bundle. Interestingly, an apoE-NT mutant in which H1′ was replaced by a short four-residue β-turn sequence shows enhanced lipid binding activity suggesting that the hydrophobic region protected by H1′ is involved in lipid interaction.34 Interestingly, one of the three tryptophanes in apoE-NT, Trp39, features a large change in SASA in apoprop relative to apowat. Moreover, this residue is located at the entrance of the cavity that accommodates buried propanol molecules close to H1′. Through its side chain rotation, Trp39 facilitates the entrance and exit of propanol molecules from the bundle interior. In agreement with these observations, near-UV CD has shown a large alteration of the signal measurements in the absence and presence of propanol suggesting a change in the tryptophane environment of the protein.15 It has been proposed that the H2-H3 connecting loop, which has been demonstrated to be a very flexible fragment together with the N-terminus of H1 and the C-terminus of H4, which are all located on the same side of the apoE-NT molecule, could be the hinge region about which the opening of the bundle occurs initiated by interactions with the lipids.35 This could take place in part through electrostatic interactions between this flexible portion of the molecule and the lipid particles as the loop contains two Glu residues. The negative charges could complement the positive charges, for instance, on phosphatidylcholine head groups of the vesicles facilitating the binding to lipid vesicles and the triggering of the opening of the helix bundle.35 However, the role of these charged residues in the lipid interaction has been questioned34 as

8736 J. Phys. Chem. B, Vol. 112, No. 29, 2008 a strong phospholipid vesicle solubilization activity is observed with the anionic phospholipid DMPG.14 The H2-H3 connecting loop does not only contain acidic residues but also contains hydrophobic ones. Interestingly, these residues, Leu82, Pro84, Val85, Leu82, and to a lesser extent Pro84, Val85, and Ala86, make persisting hydrophobic contacts with the hydrophobic portion of propanol molecules. The bundle has been suggested to open by unzipping exposing the bundle hydrophobic core. The hydrophobic interactions are thus required at a further step for the association with the opening helices. The entrance of propanol molecules in one of the two sites which is located between H2 and H3 a bit further from the loop (see Figure 6) could help in this unzipping process. The 126-183 fragment of apoE in the presence of micelleforming lipid dodecylphosphocholine was shown to form a continuous amphipathic helix by NMR.28 These authors thus raised the question about why helix 4 does not terminate at the same residue in the presence and absence of lipid. In the crystal structure of apoE-NT domain, H4 ends at Ala164.7 The amino acids around this residue, Val161-Tyr-Gln-Ala164-Gly-Ala, correspond to a typical Schellman C-capping terminal motif sequence commonly found at the end of R-helices36 in agreement with the X-ray structure. This suggests that the capping should be eliminated upon lipid binding allowing the extension of helix 4. It was proposed28 that the hydrophobic interaction required for stabilizing the C-capping motif is provided in part by long-range interaction of Val161 with Leu93. It appears that H1 and H2 and H3 and H4 remain preferentially paired during the first stage of bundle opening37 with possible subsequent reorganization of helix segments.37–39 If this view is accurate, then the hinge region, implicated in the first stage of the opening, must include the H2-H3 linking loop as previously mentioned. Opening the bundle in this manner would disrupt interactions between the H2-H3 loop residues and the end of H4. Subsequent separation of H3 and H4 would abolish long-range hydrophobic interactions between Leu93 and Val161, which contribute to stabilizing the Schellman Ccapping motif. In apoprop trajectory, one detects the disruption of the C-capping motif as observed experimentally in the presence of DPC. The Leu93-Val161 hydrophobic interactions are replaced by extensive hydrophobic interactions with the propanol molecules. The first two types of essential motions could have a role in the opening of the bundle which is observed upon lipid binding. The PCA analysis shows indeed that the first mode of largeamplitude motions involves a hinge motion opening the structure around H1′ and the facing H3-H4 loop, which is a motion that propagates along the bundle. This observed mode of motion may be related to the experiments which reveal the role of H1′ in the protection of a hydrophobic region of the bundle prone to interact with lipids.34 The finding that the second eigenvector motion is related to a type of shear motion between pairs of helices accompanied by a sway of the H2-H3 linking loop can be related to the model proposed by Weisgraber2 for the bundle opening in which loop H2-H3 serves as a hinge. Conclusions The results of the present study show that n-propanol induces structural changes in apoE-NT in agreement with experimental data. These MD simulations suggest that the nascent steps initiating the bundle opening are the penetration of propanol molecule into the protein core using hydrophobic and solventmediated interactions. These changes could depict the initial steps leading to a more developed conformational change such as the one occurring upon lipid binding. Water/n-propanol mixtures may thus be useful as a simple model system for

Ortmans and Pre´vost studying the changes that occur in apoE structure during its binding to lipid surfaces. Acknowledgment. M.P. is a Maıˆtre de Recherches at the Belgian Fonds National de la Recherche Scientifique (FNRS). References and Notes (1) Mahley, R. W. Science 1988, 240, 622. (2) Weisgraber, K. H. AdV. Protein Chem. 1994, 45, 249. (3) Wiesniewski, T.; Frangione, B. Neurosci. Lett. 1992, 135, 235. (4) Aggerbeck, L. P.; Wetterau, J. R.; Weisgraber, K. H.; Wu, C-S. C.; Lindgren, F. T. J. Biol. Chem. 1988, 263, 6249. (5) Wetterau, J. R.; Aggerbeck, L. P.; Rall, S. C., Jr.; Weisgraber, K. H. J. Biol. Chem. 1994, 263, 6240. (6) Innerarity, T. L.; Pitas, R. E.; Mahley, R. W. J. Biol. Chem. 1979, 254, 4186. (7) Wilson, C.; Wardell, M. R.; Weisgraber, K. H.; Mahley, R. W.; Agard, D. A. Science 1991, 252, 1817. (8) De Pauw, M.; Vanloo, B.; Weisgraber, K. H.; Rosseneu, M. Biochemistry 1995, 34, 10953. (9) Raussens, V.; Fisher, C. A.; Goormaghtigh, E.; Ryan, R. O.; Ruysschaert, J.-M. J. Biol. Chem. 1998, 273, 25825. (10) Peters-Libeu, C. A.; Newhouse, Y.; Hatters, D. M.; Weisgraber, K. H. J. Biol Chem. 2006, 281, 1073. (11) Jonas, A.; Steinmetz, A.; Churgay, L. J. Biol. Chem. 1993, 268, 1596. (12) Mims, M. P.; Soma, M. R.; Morrisett, J. D. Biochemistry 1990, 29, 6639. (13) Chen, G. C.; Guo, L. S. S.; Hamilton, R. L.; Gordon, V.; Richards, E. G.; Kane, J. P. Biochemistry 1984, 23, 6530. (14) Raussens, V.; Slupsky, C. M.; Ryan, R. O.; Sykes, B. D. J. Biol. Chem. 2002, 277, 29172. (15) Cle´ment-Collin, V.; Leroy, A.; Monteilhet, C.; Aggerbeck, L. P. Eur. J. Biochem. 1999, 264, 358. (16) Kitao, A; Go, N. Curr. Opin. Struct. Biol. 1999, 9, 164. (17) Berendsen, H. J.; Hayward, S. Curr. Opin. Struct. Biol. 2000, 10, 165. (18) De Groot, B. L.; Dura, X.; Mark, A. E.; Grubmu¨ller, H. J. Mol. Biol. 2001, 309, 299. (19) Garcı´a, A. E.; Harman, J. G. Protein Sci. 1996, 5, 62. (20) Vaccaro, L.; Koronakis, V.; Sansom, M. S. P. Biophys. J. 2006, 91, 558. (21) Pre´vost, M.; Kocher, J.-P. Protein Eng. 1999, 12, 475. (22) Pre´vost, M.; Ortmans, I. J. Chem. Phys. B 2001, 105. (23) Brooks, B. R.; Bruccoleri, R. E.; Olafson, B. D.; States, D. J.; Swaminathan, S.; Karplus, M. J. Comput. Chem. 1983, 4, 187. (24) MacKerell, A. D., Jr.; Bashford, D.; Bellott, M.; Dunbrack, R. L., Jr.; Evanseck, J. D.; Field, M. J.; Fischer, S.; Gao, J.; Guo, H.; Ha, S.; Joseph-McCarthy, D.; Kuchnir, L.; Kuczera, K.; Lau, F. T. K.; Mattos, C.; Michnick, S.; Ngo, T.; Nguyen, D. T.; Prodhom, B.; Reiher, W. E.; Roux, B.; Schlenkrich, M.; Smith, J. C.; Stote, R.; Straub, J.; Watanabe, M.; Wiorkiewicz-Kuczera, J.; Yin, D.; Karplus, M.; MacKerell, A. D. J. Phys. Chem. 1998, 102, 3586. (25) Jorgensen, W. L.; Chandrasekhar, J.; Madura, J. D.; Impey, R. W.; Klein, M. L. J. Chem. Phys. 1983, 79, 926. (26) Ryckaert, J.-P.; Cicotti, G.; Berendsen, H. J. J. Comput. Phys. 1977, 23, 327. (27) Lindahl, E.; Hess, B.; van der Spoel, D. J. Mol. Model. 2001, 7, 306. (28) Raussens, V; Slupsky, C. M.; Sykes, B. D.; Ryan, R. O. J. Biol. Chem. 2003, 278, 25998. (29) van Aalten, D. M. F.; Amadei, A.; Linssen, A. B. M.; Eijsink, V. G. H.; Vriend, G.; Berendsen, H. J. C. Proteins: Struct., Funct., Genet. 1995, 22, 45. (30) Barrett, C. P.; Noble, P. M. Bioinformatics 2005, 21, 3174. (31) van Aalten, D. M. F.; Findlay, J. B. C.; Amadei, A.; Berendsen, H. J. C. Protein Eng. 1995, 8, 1129. (32) Thomas, P. D.; Dill, K. A. Protein Sci. 1993, 2, 2050. (33) Khandelia, H.; Kaznessis, Y. N. J. Phys. Chem. B 2005, 109, 12990. (34) Redmond, K. A.; Murphy, C.; Narayanaswami, V.; Kiss, R. S.; Hauser, P.; Guigard, E.; Kay, C. M.; Ryan, R. O. FEBS J. 2006, 273, 558. (35) Segelke, B. W.; Forstner, M.; Knapp, M.; Trakhanov, S. D.; Parkin, S.; Newhouse, Y;M.; Bellamy, H. D.; Weisgraber, K. H.; Rupp, B. Protein Sci. 2000, 9, 886. (36) Schellman, J. A. In Protein Folding; Jaenicke, R., Ed.; Elsevier/ North Holland: New York, 1980; pp 53-61. (37) Lu, B.; Morrow, J. A.; Weisgraber, K. H. J. Biol. Chem. 2000, 275, 20775. (38) Fischer, C. A.; Ryan, R. O. J. Lipid Res. 1999, 40, 93. (39) Fischer, C. A.; Narayanaswami, V.; Ryan, R. O. J. Biol. Chem. 2000, 275, 33601.

JP8002678