Amyloid Beta Peptide Folding in Reverse Micelles - American

Department of Biochemistry & Biophysics, University of Pennsylvania Perelman ..... using NAMD 2.9, to eliminate instabilities that cause them to crash...
0 downloads 0 Views 5MB Size
Article pubs.acs.org/JACS

Amyloid Beta Peptide Folding in Reverse Micelles Gözde Eskici† and Paul H Axelsen*,‡ †

Department of Biochemistry & Biophysics, University of Pennsylvania Perelman School of Medicine, Philadelphia, Pennsylvania 19104, United States ‡ Departments of Pharmacology, Biochemistry and Biophysics, and Medicine, University of Pennsylvania, Philadelphia, Pennsylvania 19104, United States S Supporting Information *

ABSTRACT: Previously published experimental studies have suggested that when the 40-residue amyloid beta peptide is encapsulated in a reverse micelle, it folds into a structure that may nucleate amyloid fibril formation (Yeung, P. S.-W.; Axelsen, P. H. J. Am. Chem. Soc. 2012, 134, 6061). The factors that induce the formation of this structure have now been identified in a multi-microsecond simulation of the same reverse micelle system that was studied experimentally. Key features of the polypeptide−micelle interaction include the anchoring of a hydrophobic residue cluster into gaps in the reverse micelle surface, the formation of a beta turn at the anchor point that brings N- and C-terminal segments of the polypeptide into proximity, high ionic strength that promotes intramolecular hydrogen bond formation, and deformation of the reverse micelle surface to facilitate interactions with the surface along the entire length of the polypeptide. Together, these features cause the simulation-derived vibrational spectrum to red shift in a manner that reproduces the red-shift previously reported experimentally. On the basis of these findings, a new mechanism is proposed whereby membranes nucleate fibril formation and facilitate the in-register alignment of polypeptide strands that is characteristic of amyloid fibrils.



INTRODUCTION Amyloid beta (Aβ) peptides are the unstructured cleavage products of a common membrane protein in the brain. In Alzheimer’s disease, Aβ peptides undergo a conformational change, aggregate into fibrils with in-register parallel β-sheet structure, and coalesce into plaques. Since amyloid plaques become centers of oxidative stress and neuronal death, the conditions that induce Aβ peptides to form fibrils are of high interest. Fibril formation is a nucleated aggregation process1,2 that may be initiated by misfolded “seeds”.3,4 The concentration of Aβ40 (the 40-residue Aβ peptide) required to form fibrils in vitro is micromolar,2,5,6 but the concentration of Aβ peptides in the cerebrospinal fluid is subnanomolar.7 Therefore, the relevance of fibril nucleation observed in vitro at high polypeptide concentrations to in vivo fibril formation is not clear. It has been suggested that the fibril nucleation in vivo involves an interaction between Aβ peptides and lipid membranes.8 However, it is difficult to study Aβ−membrane interactions that constitute the nucleation mechanism for fibril formation when such interactions are quickly followed by fibril formation. To isolate Aβ−membrane interactions that are relevant to fibril nucleation for experimental study, Aβ40 was encapsulated in reverse micelles (RMs) under conditions where no more than one polypeptide was present in each micelle, and only ∼1% of the micelles contained a polypeptide.9 Surprisingly, the infrared amide I spectrum of the encapsulated polypeptide resembled that of a mature amyloid fibril, suggesting that one © 2017 American Chemical Society

or more aspects of the RM environment may be responsible for nucleating fibril formation. To characterize the structure of the encapsulated polypeptide, and identify the elements of an RM that determined its structure, a series of molecular dynamics simulations have now been performed on a model system that corresponds precisely to the RM system examined experimentally. Prior studies of the Aβ16−22 segment in RMs have been described in which experimental10 and computational11 results showed that water restriction within an RM enhanced polypeptide aggregation. However, the Aβ16−22 segment forms antiparallel β-sheets, in contrast to Aβ40, which has in-register parallel β-sheet structure in fibrils.12 Thus, studies of full-length Aβ are necessary to address questions about the nature of nucleation events that lead to fibril formation, in particular the process by which polypeptides are aligned in-register and parallel. In the current study, Aβ40 was encapsulated in an RM made with sodium bis(2-ethylhexyl) sulfosuccinate (AOT) suspended in isooctane (isoO), and a water-loading ratio (W0) of 11.4 (i.e., identical to that of the previously described experimental study).9 The simulation was propagated for 3 μs on the Anton supercomputer, and it revealed an array of Aβ−membrane interactions that anchor the polypeptide into the surfactant layer and induce the formation of secondary structure. A set of nine control simulations were also performed to isolate the factors responsible for the observed phenomena. The behavior Received: April 3, 2017 Published: June 20, 2017 9566

DOI: 10.1021/jacs.7b03333 J. Am. Chem. Soc. 2017, 139, 9566−9575

Article

Journal of the American Chemical Society

Figure 1. Steps in the simulation protocols for the six RM systems and four water box systems. cluster with SO3 groups oriented inward. An additional three randomly selected water molecules were replaced with sodium cations to neutralize the −3 charge of the polypeptides. Next, these polypeptide/ water/AOT systems were placed in the center of a cube with a length of 105 Å that contained 2899 molecules of isoO (nISO), a number chosen so that the isoO mass was 84% of the total system (i.e., system containing polypeptide, water, AOT, and isoO) mass and, thus, within the RM-forming portion of the AOT/water/isoO phase diagram.16 The numbers of isoO molecules in the WT and WTAlt systems was subsequently increased to 3926 and 3846, respectively, to prevent contact between RM images when the RMs became eccentric. A separate RM simulation system was constructed with the same method and composition, but with 202 fewer water molecules, corresponding to the volume of an Aβ40 peptide (∼6050 Å3). Four water box systems were also created with the VMD solvation plugin, using the coordinates of the polypeptide and a cubic shape with an edge length of 66 Å. As in the RM systems, the VMD ionization plugin was used to replace three randomly selected water molecules with sodium cations to neutralize the system. In one of the four systems, the polypeptide was unconstrained, while in a second system, 5 kcal/mol constraints were applied to polypeptide to constrain it to a 20 Å sphere to mimic RM encapsulation. In a third system, the polypeptide was unconstrained, and 3 M NaCl was used to mimic 6 M sodium cation concentration in RM. The fourth system included both the spherical constraint and 3 M NaCl. RM systems were energy minimized in six stages, each stage consisting of 0.01 ns of minimization and 1 ns of NPT simulation. In stage 1, all atoms except those in the isoO solvent were fixed in position. In stage 2, the hydrocarbon tails of the AOT anions were unfixed. In stage 3, the remaining portions of the AOT anions were unfixed. In stage 4, water and sodium cations were unfixed. In stage 5, side chains of protein were unfixed. In stage 6, all molecules in the system were unfixed. WB systems were energy minimized with only the latter three stages. Analysis. Eccentricities were calculated as described previously.24 Various VMD plugins were used for analysis. The Ramachandran plot plugin was used to characterize ϕ/ψ angles. The NAMD ENERGY plugin was used to calculate VdW interaction energies. The SASA plugin with a 1.4 Å solvent probe was used to calculate solvent accessible surface area. The IR Spectral Density Calculator plugin25 was used with default settings and a temperature of 293 K to calculate vibrational spectra in the amide I region of the infrared spectrum. This plugin requires a protein structure file and trajectory as input data and calculates a vibrational spectrum by auto-correlating the overall dipole moment of the polypeptide and performing a reverse Fourier transform. To obtain frequencies with a resolution of 2 cm−1, a trajectory saved at 1 fs intervals was required. However, the simulations performed on Anton could only be saved at 240 ps intervals due to storage limitations. To

of Aβ40 observed within an RM is distinctive, and it suggests a new mechanism whereby polypeptide strands may be aligned in-register, an essential early step in the nucleation of amyloid fibrils.



METHODS

Software, Hardware, and Parameters. All minimizations and all equilibration simulations were performed with NAMD2.913 and the CHARMM27 all-atom force field for proteins and lipids.14 Parameters for the AOT anion and isoO were obtained from previously published studies.11,15−17 Production runs were performed either on a local 32node Linux cluster using NAMD2.9 or on a 512-node Anton supercomputer for molecular dynamics.18 Systems were equilibrated as NPT ensembles, long-range electrostatic forces were calculated with the Particle Mesh Ewald method,19 an interaction cutoff of 12 Å was applied within periodic boundary conditions, and van der Waals (VdW) forces were smoothly shifted to zero between 10 and 12 Å. Equations of motion were integrated with the Verlet method and a time step of 2 fs. Langevin dynamics with a damping coefficient of 5 ps−1 was used to keep the temperature at 293 K. The pressure was maintained at 1 atm using a Nosé−Hoover−Langevin piston.20 Coordinates were saved every 0.002 ns. The systems simulated on Anton were propagated for 20 ns using NAMD 2.9 on the local 32-node Linux cluster before transferring them to the Anton machine for long time scale simulations. Simulation procedure and times are summarized in Figure 1. Anton simulations were performed at 293 K using the Nosé−Hoover thermostat and the MTK barostat. The cutoff for VdW and short-range electrostatic interactions was 9.79 Å. Long-range electrostatic interactions were computed using the Gaussian split Ewald method21 with the 64×64×64 grid size. Coordinates were saved every 0.24 ps. System Design, Equilibration, and Simulation. Aβ40 was created as extended linear chain with Pymol 1.822 and energyminimized in vacuo with NAMD. Starting structures for polypeptide with other sequences were created by mutating the side chains with Visual Molecular Dynamics (VMD).23 An alternative starting structure was prepared by creating a polypeptide with the scrambled sequence as an extended linear chain with Pymol 1.8,22 minimizing in vacuo with NAMD and mutating side chains back to the wild-type sequence with VMD. Simulation systems in which polypeptides were encapsulated in RMs were constructed with W0 = 11.4, the value used in our earlier experimental study,9 and nAOT = 127, the result of eq 7-e in our earlier theoretical study.24 The number of water molecules was nwater = W0nAOT = 1448. Polypeptides were solvated in a spherical cluster of (nwater + nAOT + 3) water molecules. AOT molecules were added by replacing nAOT randomly selected water molecules with sodium cations and distributing the anionic portions randomly on the surface of the 9567

DOI: 10.1021/jacs.7b03333 J. Am. Chem. Soc. 2017, 139, 9566−9575

Article

Journal of the American Chemical Society Table 1. System Compositionsa system name

conformation

nwater

nsod

WT WTAlt SC WTSC SCWT WT202 WB WBC WBI WBCI

1 2 1 1 1 1 1 1 1 1

1448 1448 1448 1448 1448 1246 8275 8275 7302 7302

130 130 130 130 130 130 3 3 487 487

ncl

nAOT

nISO

natoms

f iso (%)

rW (Å)

s (Å)

127 127 127 127 127 127

3926 3846 2899 2899 2899 2899

115403 113323 88701 88701 88701 88095 25426 25426 23478 23478

87 87 84 84 84 84

20 20 20 20 20 20

105 105 96 96 96 96 66 66 66 66

487 487

a

nwater, nsod, ncl, nAOT, and nISO are the numbers of water molecules, sodium cations, chloride anions, AOT, isooctane molecules; natoms is the total number of atoms in each system. f iso is the mass fraction of isoO as a percentage of mass of all components in the system.rW is the average distance from the center of the RM to the outermost water molecule. s is the edge length of the unit cell.

Table 2. Polypeptide Sequences name WT SC WTSC SCWT

sequence DAEFR KVKGL DAEFR YVKGE

HDSGY IDGDH HDSGY IDGDH

EVHHQ IGDLV EVHHF IGDLQ

KLVFF YEFMA KAVQL KLVFF

AEDVG SNSFR FEDVG ANSFR

SNKGA EGVGA SNKGA EGVGA

IIGLM GHVHV IIGLM GHVHV

VGGVV AQVEF VGGVV AMVES

Figure 2. RM eccentricity. Upper panel: eccentricity of the WT system over 3000 ns. Lower panels: eccentricity of the WTAlt, SC, WTSC, SCWT, and WT202 systems over 500 ns. create trajectories with shorter step intervals for spectral analysis, seven snapshots were selected with at least 50 ns intervals between them from equilibrated Anton simulations, and the polypeptide coordinates were excerpted from these snapshots. The motions of the main chain atoms were simulated in vacuo for 16 ps using NAMD with side chain atom positions fixed to prevent structural distortion. To correct for the ∼50 cm−1 blue shift that invariably occurs when predicting infrared spectra from molecular mechanics simulations, as well as the use of D2O in the experimental studies, all frequencies resulting from this procedure were uniformly scaled by 0.97. The intrinsic line width of each frequency was modeled as a Lorentzian distribution with a full width at half-maximum of 8 cm−1, and the distributions were summed to yield a spectrum. Finally, spectra derived in this manner from the seven snapshots were averaged, and the maximum amplitudes were normalized.

alternative backbone conformation (WTAlt) (Table 2). A third RM system (SC) had the same scrambled amino acid sequence used in previously published experimental work,9 while only residues 15−21 were scrambled in a fourth system (WTSC). A fifth system was identical to SC, except that the wild-type sequence of residues 15−21 was restored (SCWT). The sixth system was an identical replica of WT except that a volume of water equivalent to the volume of the polypeptide was removed (WT202). Each of the six RM systems began with a single polypeptide strand bearing no recognizable secondary structure or intramolecular hydrogen bonding, positioned at the center of a spherical RM with a radius of 20 Å and all residues at least 5 Å from the surfactant. The four water box systems included one with Aβ40 as a random coil in the center of a water box (WB). A second system was identical to WB except that the polypeptide was restrained in a soft but non-deformable 20 Å sphere (WBC). A third water box system had no restraints but included 3 M sodium chloride to mimic the high internal ionic strength of the WT system (WBI). The fourth water box system included both the spherical constraint and 3 M sodium chloride (WBCI).



RESULTS Initial Structures and Conditioning. The procedures described above yielded 10 simulation systems: six RM and four water box systems (Table 1). These included one RM system containing the wild-type Aβ40 sequence (WT), and a second system that was an identical replica of WT except for an 9568

DOI: 10.1021/jacs.7b03333 J. Am. Chem. Soc. 2017, 139, 9566−9575

Article

Journal of the American Chemical Society After energy minimization, RM systems were conditioned by simulating them as NPT ensembles for 20 ns using NAMD 2.9, to eliminate instabilities that cause them to crash with “momentum exceeded” errors on Anton (Figure 1). During the 20 ns conditioning period, the shape of each RM became markedly non-spherical, although all portions of the polypeptide remained at least 3 Å from the surfactant. The four water box systems were not run on Anton, so they did not require conditioning, but they were conditioned for 20 ns anyway to facilitate comparison with the RM systems. During conditioning, the edge length of the water box systems shortened to 62.5 Å. In WBC and WBCI, all polypeptide residues remained at least 0.5 Å from the restraint limit. Equilibration. The overall system energies and volumes in all 10 simulation systems reached apparent equilibrium values during the 20 ns conditioning periods. During subsequent simulations, there was no net water gain or loss from the RMs, and water molecules were only rarely observed in the isoO phase. Therefore, the energies, volumes, and water contents of each RM system appeared to be well equilibrated at the end of the 20 ns conditioning periods. However, the WT and SC systems required approximately 200 ns for eccentricity values to plateau, while the eccentricity values of the other systems appeared to require less than 100 ns (Figure 2). The evolution of polypeptide conformation within each system was evaluated (Figure S1), and almost every secondary structure that eventually formed in any system had formed within 200 ns. Therefore, we concluded that all systems were suitably equilibrated by 200 ns, and subsequent analyses were only performed on portions of the simulations beyond 200 ns. Polypeptide Interactions with RM Surface. In each RM system, gaps large enough to pass a water molecule were frequently observed in the RM surface between surfactant molecules (Figure 3). Some of these gaps were filled by isoO, permitting direct contact between isoO molecules and components of the RM core. Other gaps were filled by hydrophobic amino acid side chains, which consequently made direct contact with both the isoO and the hydrocarbon tails of the surfactant. There are three possible explanations for the occurrence of these surface gaps. One is that they form because deviations from a spherical shape increase the surface-area-to-volume ratio to a point where AOT molecules cannot cover the surface. A second possible explanation is that polypeptide encapsulation causes the RM volume to expand to a point where AOT molecules cannot cover the surface. A third possible explanation is that side chain insertion into the surface is energetically favorable and thereby creates gaps that would not otherwise exist. These possibilities are not mutually exclusive. To evaluate these possibilities, SASAs were determined for each structure snapshot saved from the last 100 ns of each of the six simulation systems, as well as from an RM simulation identical to the WT system except without encapsulated polypeptide (designated RM127 because all of these RM systems have 127 AOT molecules). The total SASA, and individual contributions of water, sodium ions, protein, and AOT to the SASA, were averaged separately (Figure 4). The “gap area” was defined as the sum of the SASAs for water, sodium ions, and protein, and is represented as a percentage of total SASA. Results show that the polypeptide-free RM127 system and the polypeptide-containing WT202 systems had the smallest gap areas. Because the only difference between these two systems was the presence of polypeptide, and waters were removed to

Figure 3. Structure snapshot from the WT system showing gaps in the surfactant coverage of the surface and the occupation of these gaps by hydrophobic residue side chains. Color code: violet surface, AOT headgroups (SO3− only; a 1.4 Å probe was used to calculate the surface); cyan spheres, water; red tubes, polypeptide.

compensate for polypeptide volume in the WT202 system, it appears that gaps form due to shape distortion of the RM, irrespective of whether an encapsulated polypeptide was present. In other simulation systems, where water molecules were not removed to compensate for the added volume of the encapsulated polypeptide, gap areas were invariably larger. These results show that the expansion of RM volume by polypeptide encapsulation also increased gap area. Although the WT202 and RM127 systems had similar gap areas as a percentage of total SASA, the total SASA for RM127 was much larger than the total SASA for WT202. The most significant contributions to the larger SASA of RM127 were due to the SASAs of AOT molecules and waters. Since the number of AOT molecules was the same in both systems, the larger AOT area must be due to a larger portion of AOT molecules being solvent exposed, either through a change in their shape, or because additional aspects of AOT molecules become accessible to the 1.4 Å surface area probe. In the other five polypeptide-containing RM systems (WT, WTAlt, WTSC, SC, and SCWT) gap areas varied with the total SASA, as would be expected in systems with the same fixed number of AOT molecules. To quantify the number and size of the gap areas, and the extent to which amino acid side chains occupied them, VdW interaction energies were calculated between each amino acid side chain and the set of all isoO molecules (Figure 5). Hydrophobic side chains in the WT, WTAlt, and WT202 systems tended to have larger interaction energies with isoO than in other systems. Visual inspection of these simulations confirmed that the side chains of Leu17, Phe19, and Phe20 typically situated in the gaps between surfactant molecules, making contact with AOT tails and isoO molecules. Overall interaction energies were markedly smaller in the SC and WTSC systems, 9569

DOI: 10.1021/jacs.7b03333 J. Am. Chem. Soc. 2017, 139, 9566−9575

Article

Journal of the American Chemical Society

RMs. The polypeptides in the WT, WTAlt, WT202, and SCWT systems tended to associate with the surfactant and distort the RM shape to tightly envelop them. The overall closer proximity between polypeptide and surfactant headgroups is reflected in the larger numbers of sulfate groups within 3 Å of the polypeptides in these systems (Figure S2). The SC and WTSC polypeptides, in contrast, assumed central positions within the RM with relatively few sulfate groups in their vicinity. Differences in backbone torsion angles between the WT and SC systems are shown as 3D Ramachandran plots in Figure 6. The most prominent peak in both systems is located near a region of the plot that is characteristic of α-helices. However, these peaks arise from isolated individual residues, not helices, with torsion angles that are stabilized by interactions with the side chains of adjacent residues. The most conspicuous difference between the two systems is a large peak in the WT system in a region characteristic of β-sheets. The significance of this difference is addressed in the discussion below. Another prominent feature of the WT system is a β turn involving residues 18−21 that persisted throughout the 3 μs simulation (Figure 7). This turn is also evident in the WTAlt, WT202, and SCWT systems, but not in the SC and WTSC systems, indicating that the clustering of residues 18−21 in sequence was required for β turn formation. In view of this observation, it should be recalled that the side chains of the residues involved in this turn are the same side chains that tended to insert into gaps between surfactant molecules on the RM surface (Figures 3 and 5). The polypeptides in WB systems did not exhibit any persistent secondary structure aside from a short segment of αhelix in WBC involving residues 11−16, and several short stretches of extended β-sheet formed in WBI. The WBCI system formed various turns involving residues 17−25, but not the β turn observed in the WT system involving residues 18−21. Therefore, although confinement and high ionic strength promote the formation of turns, interactions with surfactant are required to form the stable β turn at residues 18−21. In sum, Aβ40 within a reverse micelle appears to assume a hairpin shape with a stable β turn, with hydrophobic side chains projecting from the end of the hairpin into the surfactant, and with the remainder of the polypeptide tightly enveloped by surfactant (Figure 8). Vibrational Spectra. Vibrational spectra were calculated for the amide I vibration of the polypeptide in the WT, WTAlt, WT202, and SC systems and compared to that of the initial random coil (Figure 9). The SC and random coil spectra were broad with maxima at 1640−1643 cm−1. However, spectral maxima for the WT, WTAlt, and WT202 systems were shifted to 1628−1631 cm−1, consistent with the previously published experimental IR spectra obtained from Aβ40 encapsulated in an RM.9,26 Selective removal of residues from the spectral calculation showed that the shifted spectrum arose from the N-terminal 18 residues of Aβ40 (Figure 8).

Figure 4. Solvent-accessible surface areas (SASAs) of system components and gap areas in the surfactant layer (i.e., the sum of the SASAs for water, sodium ions, and protein) for RM systems as a percentage of total SASA. Error bars represent standard deviations for the distribution of results.

but the relatively large interaction energies between Phe19, Phe20, and isoO were restored in the SCWT system. These observations suggest that the chemical context of the polypeptide side chains (i.e., their sequences) influenced the extent to which hydrophobic side chains occupied gaps in the surfactant layer. More specifically, this observation suggested that the cluster of hydrophobic side chains among residues 15− 21 promoted their association with gaps. We conclude that gap formation occurs regardless of polypeptide presence, but that an encapsulated polypeptide increases gap area. The occupation of surface gaps by hydrophobic side chains appears to be energetically favorable, particularly when they are clustered together in sequence. However, the energetics and/or extent of gap occupation by hydrophobic side chains was insufficient to cause a measurable increase in gap formation. Polypeptide Orientation and Conformation. Visual inspection of the simulation trajectories revealed significant differences in the way that polypeptides situated within the



DISCUSSION This investigation provides detailed insight into previously published experimental observations about the remarkable behavior of Aβ40 peptides encapsulated in RMs. The main experimental observation was that monomeric Aβ40 in a RM had a red-shifted amide I spectrum, similar to that of fibrillar Aβ40, suggesting that the RM environment had induced monomeric Aβ40 to form a structure capable of nucleating amyloid fibril formation. The most striking result of the current 9570

DOI: 10.1021/jacs.7b03333 J. Am. Chem. Soc. 2017, 139, 9566−9575

Article

Journal of the American Chemical Society

Figure 5. Average van der Waals interaction energies between each side chain and the set of all isoO molecules for each residue in the six RM systems. Error bars represent standard deviations for the distribution of results.

close and extensive interaction between polypeptide and membrane surface (discussed as “a third type of interaction”, below). Most of the N-terminal 18 residues in the WT simulation (see Figure 8) contribute to a narrow band of backbone torsion angles in the β sheet region of the Ramachandran plot (Figure 6, left). In molecular mechanics simulations, spectral shifts must be inferred from the orientation, motions, and nonbonded interactions of fixed partial charges, which were characterized in the present study by the analysis of fluctuations in the magnitude of the overall macrodipole.25 This approach is unable to predict the kind of splitting that gives rise to the highfrequency component often observed in model compounds that form antiparallel β sheets. To predict high-frequency components that arise from this kind of splitting amide I splitting, more sophisticated analyses based on transition dipole or exciton coupling are required.11,31,38 However, these analyses

study is that simulations have reproduced the experimentally observed red-shift in the amide I vibrational spectrum. Shifts of this nature are due to the alignment and coupling of oscillating amide I dipoles, chiefly the N−CO groups in the polypeptide backbone. Coupling among these groups to yield red-shifted vibrational spectra is a well-known feature of parallel and antiparallel β sheets,27−37 and it arises as a consequence of polypeptide backbone geometry and the torsion angle distribution in a β sheet that is stabilized by its interstrand hydrogen-bonding pattern. Such is the case in a mature amyloid fibril. There were no β sheets observed in any of the simulated RM systems described above, yet the amide I dipoles in the WT systems are aligned in the manner found in β sheets, as a consequence of a torsion angle distribution that is also characteristic of β sheets (Figure 6, left). In the case of the WT system this torsion angle distribution is stabilized by the 9571

DOI: 10.1021/jacs.7b03333 J. Am. Chem. Soc. 2017, 139, 9566−9575

Article

Journal of the American Chemical Society

Figure 6. 3D Ramachandran plots for the WT and SC systems. The vertical (z) axis has been normalized so that dark blue is zero, while the maximum is red. The peak at ϕ = −60, ψ = −40 in both plots arises from isolated single residues that exhibit limited backbone dihedral rotations due to side-chain interactions; they are not α-helical despite being close to the α-helical region of the plot. The principle difference between the plots is at ϕ = −60, ψ = 170, where the WT system has a large peak, and the SC system has a small peak. These peaks are in a region characteristic of β-sheets.

Phe19 and Phe20 for folding,43 fibril formation,44−48 dimer formation,49 and lipid−membrane interactions.50 A second interaction is between the polypeptide and ions in the RM interior. By competing for available water in the hydration shells, ions tend to dehydrate the polypeptide and promote the formation of intramolecular hydrogen bonds (Figure 7).10 This conclusion is consistent with spectroscopic studies showing that high ionic concentration induces a transition from random coil to β-sheet in Aβ.51 A third type of interaction results in deformation of the RM surface by the polypeptide. Just as interactions with the RM surface alter polypeptide properties, the polypeptide will alter the RM surface.52 In this case, the anchoring of multiple hydrophobic side chains in the RM surface (Figures 3 and 8) induces surface deformation, which in turn facilitates more extensive contact between polypeptide and the RM surface than would otherwise be possible if the surface was rigid and spherical. These interactions cause the RM surface to tightly envelope the peptide, which explains why two aspects of the RM environmentphysical confinement and crowdingdid not affect Aβ behavior when applied in the WB systems. In the WT systems, configurational entropy of the polypeptide is reduced primarily by the close proximity of the RM surface all along the polypeptide chain, which confines the polypeptide to a much greater degree than it would have been confined in a spherical RM. Indeed, the polypeptide occupies only a small portion of the RM interior. This conclusion is supported by the experimental finding that Aβ behavior is similar in an RM with W0 = 11.4 or W0 = 30.9 Overall, the forgoing results are in excellent agreement with a recent but more limited NMR and simulation study of the Aβ(16−28) segment, which also found that the hydrophobic cluster spanning residues 17−21 mediates the association of this segment with membranes.53 Most previously published simulation studies of Aβ−membrane interactions, however, have been performed with simplified polypeptide models that cannot mimic the features critical to the findings of this study, such as hydrogen bonds and aromatic ring interactions,54,55 while others have been performed with simplified55−60 or heterogeneous61−64 membrane models. Still other simulation studies begin with pre-inserted polypeptide57−59,65 or multiple polypeptides60 to study mechanisms of oligomerization and

Figure 7. Stereo view (wall-eyed) of the persistent β turn observed in the WT system, showing the edge−face interaction of the two Phe residue side chains.

yield only modest improvements in the accuracy of the fundamental absorption band, and they still depend on molecular mechanical simulations to create representative structure ensembles. The origin of the high-frequency band in previously published experimental data9 may have been true antiparallel β-sheet formation, interactions between the peptide and the surfactant, or residual trifluoroacetic acid.39 The latter is a known and common interfering compound, and may have persisted despite steps taken to remove it that are usually reliable. In any case, the high-frequency band is well-resolved from the fundamental amide I band. The insight provided by the current investigation shows that Aβ40 and components of the RM interact in three distinctly different ways. One interaction is between hydrophobic portions of the surfactant layer and a hydrophobic cluster composed of residues 17−21. In particular, the side chains of Phe19 and Phe20 insert into gaps between AOT molecules, thereby anchoring the polypeptide in the RM surface through hydrophobic interactions. Steric interactions between the Phe side chains then promote β turn formation, while electrostatic interactions between their π-electron systems stabilize an edgeto-face interaction (Figure 7).40,41 This arrangement of aromatic rings and β turn formation is in excellent agreement with previously published NMR results,42 and with numerous experimental studies of Aβ peptides showing the importance of 9572

DOI: 10.1021/jacs.7b03333 J. Am. Chem. Soc. 2017, 139, 9566−9575

Article

Journal of the American Chemical Society

Figure 8. Stereo view (wall-eyed) of a typical structure from the WT system, showing AOT headgroups (SO32− only) as yellow/red spheres and rods, and blue tubes as the polypeptide main chain with selected side chains. The labeled residues each have hydrophobic side chains embedded in the surfactant layer. Regions labeled with dotted arcs are responsible for anchoring the peptide in the surfactant layer (the embedded β turn), for the red-shifted amide I spectrum (N-terminal coupled amide I oscillators), and for the unshifted amide I spectrum (C-terminal random coil).

polypeptide−membrane interactions likely to be involved in nucleation. These results are significant to questions about how Aβ peptides are induced to form amyloid fibrils in Alzheimer’s disease. Aβ peptides are normally present in human brain, but they are neither aggregated nor structured. Their conversion into an amyloid fibril with highly regular structure is a nucleated process that is promoted by interactions with a membrane.67−71 The interactions responsible for nucleation are difficult to study experimentally because they are obscured by the aggregation that ensues, whereas RM encapsulation prevents aggregation and makes interaction of the monomeric polypeptide with a membrane experimentally accessible. Several pathways and intermediates have been proposed by which monomeric polypeptides may aggregate to form fibrils.9 There are two key steps in these pathways. One is the alignment of polypeptide chains so that the β sheets are inregister. It was originally proposed that this alignment might be accomplished by the coordination of a copper ion to the His residues at positions 13 and 14 in two different Aβ peptides. The current study shows that alignment may also be accomplished by the anchoring of Phe19-Phe20 residues from two different Aβ peptides in the same membrane. The second key step is rotation of polypeptide main chains to create the intermolecular hydrogen bonds characteristic of amyloid fibrils, rather than the intramolecular hydrogen bonds observed in these simulations. These rotations need not occur all at once, especially given the experimental evidence that fibril structure can undergo significant maturation over long times.72 However, chain alignment must precede these rotations, or the energetic barriers to fibril nucleation are likely to be insurmountable. It should be noted that the modification of the three His residues in Aβ40 by 4-hydroxy-2-nonenal (HNE) is a potent way to promote fibril formation.73 HNE-modification of the His residues increases the affinity of Aβ peptides for

Figure 9. Calculated vibrational spectra for five RM systems.

aggregation. While multiple simulation studies have approached questions about fibril nucleation with a finite number of interacting polypeptides,66 the advantage of the current study is that it focuses on Aβ−membrane interactions under simulation conditions that precisely mimic the conditions that have been characterized experimentally, and that isolate the initial 9573

DOI: 10.1021/jacs.7b03333 J. Am. Chem. Soc. 2017, 139, 9566−9575

Article

Journal of the American Chemical Society membranes, and dramatically reduces the time × concentration product required for fibril formation. The mechanism by which HNE modification promotes fibril nucleation may involve the anchoring of two adjacent residue side chains in a membrane, similar to the mechanism involving Phe residues proposed herein. It would be appropriate to extend these investigations by creating RM systems containing two or more polypeptides with the aim of observing how the polypeptides interact and align. Although this study has focused on possible mechanisms of fibril nucleation, another possible outcome of multiple-peptide studies is deeper membrane penetration by the polypeptide and ion channel formation. The formation of aberrant channels is the subject of an extensive literature and may be a mechanism of neurotoxicity operating in Alzheimer’s disease.74−83 However, channel formation should be pursued with a full bilayer membrane model, rather than the surfactant monolayer model employed herein.



(9) Yeung, P. S. W.; Axelsen, P. H. J. Am. Chem. Soc. 2012, 134, 6061−6063. (10) Mukherjee, S.; Chowdhury, P.; Gai, F. J. Phys. Chem. B 2009, 113, 531−535. (11) Martinez, A. V.; Malolepsza, E.; Rivera, E.; Lu, Q.; Straub, J. E. J. Chem. Phys. 2014, 141, 22D530-1−22D530-9. (12) Antzutkin, O. N.; Balbach, J. J.; Leapman, R. D.; Rizzo, N. W.; Reed, J.; Tycko, R. Proc. Natl. Acad. Sci. U. S. A. 2000, 97, 13045− 13050. (13) Phillips, J. C.; Braun, R.; Wang, W.; Gumbart, J.; Tajkhorshid, E.; Villa, E.; Chipot, C.; Skeel, R. D.; Kale, L.; Schulten, K. J. Comput. Chem. 2005, 26, 1781−1802. (14) Foloppe, N.; MacKerell, A. D., Jr. J. Comput. Chem. 2000, 21, 86−104. (15) Abel, S.; Sterpone, F.; Bandyopadhyay, S.; Marchi, M. J. Phys. Chem. B 2004, 108, 19458−19466. (16) Abel, S.; Waks, M.; Marchi, M. Eur. Phys. J. E: Soft Matter Biol. Phys. 2010, 32, 399−409. (17) Tian, J. H.; Garcia, A. E. J. Chem. Phys. 2011, 134, 225101−1− 225101−11. (18) Shaw, D. E.; Dror, R. O.; Salmon, J. K.; Grossman, J. P. In Millisecond-Scale Molecular Dynamics Simulations on Anton. Proceedings of the ACM/IEEE Conference on Supercomputing (SC09), Portland, OR, Nov 20, 2009; Mackenzie, K. M., Banka, J. A., Eds.; IEEE Press: Piscataway, NJ, 2009 (19) Darden, T.; York, D.; Pedersen, L. J. Chem. Phys. 1993, 98, 10089−10092. (20) Feller, S. E.; Zhang, Y.; Pastor, R. W.; Brooks, B. R. J. Chem. Phys. 1995, 103, 4613−4621. (21) Shan, Y. B.; Klepeis, J. L.; Eastwood, M. P.; Dror, R. O.; Shaw, D. E. J. Chem. Phys. 2005, 122, 054101−1−054101−13. (22) The PyMOL Molecular Graphics System, Version 1.8; Schrödinger, LLC, 2016. (23) Humphrey, W.; Dalke, A.; Schulten, K. J. Mol. Graphics 1996, 14, 33−38. (24) Eskici, G.; Axelsen, P. H. J. Phys. Chem. B 2016, 120, 11337− 11347. (25) Kohlmeyer, A. VMD IR-spec plugins, Release 1.3; Zenodo, 2017; DOI: 10.5281/zenodo.556193. (26) Yeung, P. S. W.; Eskici, G.; Axelsen, P. H. Biochim. Biophys. Acta, Biomembr. 2013, 1828, 2314−2318. (27) Miyazawa, T.; Blout, E. R. J. Am. Chem. Soc. 1961, 83, 712−719. (28) Kubelka, J.; Keiderling, T. A. J. Am. Chem. Soc. 2001, 123, 12048−12058. (29) Kubelka, J.; Keiderling, T. A. J. Am. Chem. Soc. 2001, 123, 6142−6150. (30) Lee, C.; Cho, M. H. J. Phys. Chem. B 2004, 108, 20397−20407. (31) Paul, C.; Wang, J. P.; Wimley, W. C.; Hochstrasser, R. M.; Axelsen, P. H. J. Am. Chem. Soc. 2004, 126, 5843−5850. (32) Paul, C.; Axelsen, P. H. J. Am. Chem. Soc. 2005, 127, 5754− 5755. (33) Londergan, C. H.; Wang, J. P.; Axelsen, P. H.; Hochstrasser, R. M. Biophys. J. 2006, 90, 4672−4685. (34) Kim, Y. S.; Liu, L.; Axelsen, P. H.; Hochstrasser, R. M. Proc. Natl. Acad. Sci. U. S. A. 2008, 105, 7720−7725. (35) Kim, Y. S.; Liu, L.; Axelsen, P. H.; Hochstrasser, R. M. Proc. Natl. Acad. Sci. U. S. A. 2009, 106, 17751−17756. (36) Shim, S. H.; Gupta, R.; Ling, Y. L.; Strasfeld, D. B.; Raleigh, D. P.; Zanni, M. T. Proc. Natl. Acad. Sci. U. S. A. 2009, 106, 6614−6619. (37) Buchanan, L. E.; Carr, J. K.; Fluitt, A. M.; Hoganson, A. J.; Moran, S. D.; de Pablo, J. J.; Skinner, J. L.; Zanni, M. T. Proc. Natl. Acad. Sci. U. S. A. 2014, 111, 5796−5801. (38) Wang, L.; Middleton, C. T.; Zanni, M. T.; Skinner, J. L. J. Phys. Chem. B 2011, 115, 3713−3724. (39) Surewicz, W. K.; Mantsch, H. H. J. Mol. Struct. 1989, 214, 143− 147. (40) Burley, S. K.; Petsko, G. A. FEBS Lett. 1986, 203, 139−143. (41) Burley, S. K.; Petsko, G. A. Science 1985, 229, 23−28.

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/jacs.7b03333. Additional data and figures (PDF)



AUTHOR INFORMATION

Corresponding Author

*[email protected] ORCID

Paul H Axelsen: 0000-0002-7118-1641 Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This work was supported by grants from the NIH (GM76201) and the Alzheimer’s Association (to P.H.A.). It made use of the Extreme Science and Engineering Discovery Environment (XSEDE), supported by National Science Foundation grant number OCI-1053575, and the Bridges system at the Pittsburgh Supercomputing Center (PSC) supported by NSF award number ACI-1445606. Access to the Anton machine was made possible by the National Center for Multiscale Modeling of Biological Systems through grant number MCB150023P from the Pittsburgh Supercomputing Center. Görkem Eskici assisted with figure design.



REFERENCES

(1) Fezoui, Y.; Teplow, D. B. J. Biol. Chem. 2002, 277, 36948−36954. (2) Lomakin, A.; Chung, D. S.; Benedek, G. B.; Kirschner, D. A.; Teplow, D. B. Proc. Natl. Acad. Sci. U. S. A. 1996, 93, 1125−1129. (3) Terzi, E.; Holzemann, G.; Seelig, J. Biochemistry 1997, 36, 14845−14852. (4) Kirkitadze, M. D.; Condron, M. M.; Teplow, D. B. J. Mol. Biol. 2001, 312, 1103−1119. (5) Sabate, R.; Estelrich, J. J. Phys. Chem. B 2005, 109, 11027−11032. (6) Soreghan, B.; Kosmoski, J.; Glabe, C. J. Biol. Chem. 1994, 269, 28551−28554. (7) Seubert, P.; Vigo-Pelfrey, C.; Esch, F.; Lee, M.; Dovey, H.; Davis, D.; Sinha, S.; Schlossmacher, M.; Whaley, J.; Swindlehurst, C.; et al. Nature 1992, 359, 325−327. (8) Axelsen, P. H.; Komatsu, H.; Murray, I. V. J. Physiology 2011, 26, 54−69. 9574

DOI: 10.1021/jacs.7b03333 J. Am. Chem. Soc. 2017, 139, 9566−9575

Article

Journal of the American Chemical Society

(78) Jang, H.; Zheng, J.; Nussinov, R. Biophys. J. 2007, 93, 1938− 1949. (79) Jang, H.; Arce, F. T.; Capone, R.; Ramachandran, S.; Lal, R.; Nussinov, R. Biophys. J. 2009, 97, 3029−3037. (80) Do, T. D.; LaPointe, N. E.; Nelson, R.; Krotee, P.; Hayden, E. Y.; Ulrich, B.; Quan, S.; Feinstein, S. C.; Teplow, D. B.; Eisenberg, D.; Shea, J. E.; Bowers, M. T. J. Am. Chem. Soc. 2016, 138, 549−557. (81) Jang, H.; Arce, F. T.; Ramachandran, S.; Capone, R.; Azimova, R.; Kagan, B. L.; Nussinov, R.; Lal, R. Proc. Natl. Acad. Sci. U. S. A. 2010, 107, 6538−6543. (82) Jang, H.; Arce, F. T.; Ramachandran, S.; Capone, R.; Lal, R.; Nussinov, R. J. Phys. Chem. B 2010, 114, 9445−9451. (83) Jang, H.; Arce, F. T.; Ramachandran, S.; Capone, R.; Lal, R.; Nussinov, R. J. Mol. Biol. 2010, 404, 917−934.

(42) Zhang, S.; Iwata, K.; Lachenmann, M. J.; Peng, J. W.; Li, S.; Stimson, E. R.; Lu, Y.; Felix, A. M.; Maggio, J. E.; Lee, J. P. J. Struct. Biol. 2000, 130, 130−141. (43) Sikorski, P.; Atkins, E. D. T.; Serpell, L. C. Structure 2003, 11, 915−926. (44) Esler, W. P.; Stimson, E. R.; Ghilardi, J. R.; Lu, Y. A.; Felix, A. M.; Vinters, H. V.; Mantyh, P. W.; Lee, J. P.; Maggio, J. E. Biochemistry 1996, 35, 13914−13921. (45) Wurth, C.; Guimard, N. K.; Hecht, M. H. J. Mol. Biol. 2002, 319, 1279−1290. (46) Makin, O. S.; Atkins, E.; Sikorski, P.; Johansson, J.; Serpell, L. C. Proc. Natl. Acad. Sci. U. S. A. 2005, 102, 315−320. (47) Cukalevski, R.; Boland, B.; Frohm, B.; Thulin, E.; Walsh, D.; Linse, S. ACS Chem. Neurosci. 2012, 3, 1008−1016. (48) Bernstein, S. L.; Wyttenbach, T.; Baumketner, A.; Shea, J. E.; Bitan, G.; Teplow, D. B.; Bowers, M. T. J. Am. Chem. Soc. 2005, 127, 2075−2084. (49) Gnanakaran, S.; Nussinov, R.; Garcia, A. E. J. Am. Chem. Soc. 2006, 128, 2158−2159. (50) Di Scala, C.; Yahi, N.; Lelievre, C.; Garmy, N.; Chahinian, H.; Fantini, J. ACS Chem. Neurosci. 2013, 4, 509−517. (51) Jarvet, J.; Damberg, P.; Bodell, K.; Eriksson, L. E. G.; Graslund, A. J. Am. Chem. Soc. 2000, 122, 4261−4268. (52) Morriss-Andrews, A.; Brown, F. L. H.; Shea, J. E. J. Phys. Chem. B 2014, 118, 8420−8432. (53) Bera, S.; Korshavn, K. J.; Kar, R. K.; Lim, M. H.; Ramamoorthy, A.; Bhunia, A. Phys. Chem. Chem. Phys. 2016, 18, 16890−16901. (54) Davis, C. H.; Berkowitz, M. L. Biophys. J. 2009, 96, 785−797. (55) Davis, C. H.; Berkowitz, M. L. J. Phys. Chem. B 2009, 113, 14480−14486. (56) Mobley, D. L.; Cox, D. L.; Singh, R. R. P.; Maddox, M. W.; Longo, M. L. Biophys. J. 2004, 86, 3585−3597. (57) Xu, Y. C.; Shen, J. J.; Luo, X. M.; Zhu, W. L.; Chen, K. X.; Ma, J. P.; Jiang, H. L. Proc. Natl. Acad. Sci. U. S. A. 2005, 102, 5403−5407. (58) Lemkul, J. A.; Bevan, D. R. Protein Sci. 2011, 20, 1530−1545. (59) Lemkul, J. A.; Bevan, D. R. FEBS J. 2009, 276, 3060−3075. (60) Brown, A. M.; Bevan, D. R. Biophys. J. 2016, 111, 937−949. (61) Manna, M.; Mukhopadhyay, C. PLoS One 2013, 8, e71308. (62) Zhou, X. L.; Xu, J. PLoS One 2012, 7, e46245. (63) Lemkul, J. A.; Bevan, D. R. Biochemistry 2013, 52, 4971−4980. (64) Zhao, L. N.; Chiu, S. W.; Benoit, J.; Chew, L. Y.; Mu, Y. G. J. Phys. Chem. B 2011, 115, 12247−12256. (65) Lemkul, J. A.; Bevan, D. R. Arch. Biochem. Biophys. 2008, 470, 54−63. (66) Shea, J. E.; Urbanc, B. Curr. Top. Med. Chem. 2013, 12, 2596− 2610. (67) McLaurin, J.; Chakrabartty, A. Eur. J. Biochem. 1997, 245, 355− 363. (68) Terzi, E.; Holzemann, G.; Seelig, J. J. Mol. Biol. 1995, 252, 633− 642. (69) Sabate, R.; Gallardo, M.; Estelrich, J. Colloids Surf., A 2005, 270271, 13−17. (70) Kremer, J. J.; Pallitto, M. M.; Sklansky, D. J.; Murphy, R. M. Biochemistry 2000, 39, 10309−10318. (71) Kremer, J. J.; Sklansky, D. J.; Murphy, R. M. Biochemistry 2001, 40, 8563−8571. (72) Ma, J. Q.; Komatsu, H.; Kim, Y. S.; Liu, L.; Hochstrasser, R. M.; Axelsen, P. H. ACS Chem. Neurosci. 2013, 4, 1236−1243. (73) Liu, L.; Komatsu, H.; Murray, I. V. J.; Axelsen, P. H. J. Mol. Biol. 2008, 377, 1236−1250. (74) Arispe, N.; Rojas, E.; Pollard, H. B. Proc. Natl. Acad. Sci. U. S. A. 1993, 90, 567−571. (75) Arispe, N.; Pollard, H. B.; Rojas, E. Proc. Natl. Acad. Sci. U. S. A. 1993, 90, 10573−10577. (76) Quist, A.; Doudevski, L.; Lin, H.; Azimova, R.; Ng, D.; Frangione, B.; Kagan, B.; Ghiso, J.; Lal, R. Proc. Natl. Acad. Sci. U. S. A. 2005, 102, 10427−10432. (77) Connelly, L.; Jang, H.; Arce, F. T.; Ramachandran, S.; Kagan, B. L.; Nussinov, R.; Lal, R. Biochemistry 2012, 51, 3031−3038. 9575

DOI: 10.1021/jacs.7b03333 J. Am. Chem. Soc. 2017, 139, 9566−9575