doi:10.1016/j.jmb.2011.12.063

J. Mol. Biol. (2012) 421, 572–586 Contents lists available at www.sciencedirect.com

Journal of Molecular Biology j o u r n a l h o m e p a g e : h t t p : / / e e s . e l s e v i e r. c o m . j m b

Molecular Interactions of Alzheimer's Aβ Protofilaments with Lipid Membranes Florentina Tofoleanu and Nicolae-Viorel Buchete⁎ School of Physics, University College Dublin, Belfield, Dublin 4, Ireland Complex and Adaptive Systems Laboratory, University College Dublin, Belfield, Dublin 4, Ireland Received 23 October 2011; received in revised form 22 December 2011; accepted 29 December 2011 Available online 17 January 2012 Edited by S. Radford Keywords: Alzheimer's amyloid β peptide fibrils; amyloid protofilament structure; molecular dynamics simulations; amyloid peptide–lipid membrane interactions; toxic amyloid channels

Amyloid fibrils and peptide oligomers play central roles in the pathology of Alzheimer's disease, type 2 diabetes, Parkinson's disease, Huntington's disease, and prion-related disease. Here, we investigate the molecular interactions between preformed amyloid β (Aβ) molecular protofilaments and lipid bilayer membranes, in the presence of explicit water molecules, using computational models and all-atom molecular dynamics. These interactions play an important role in the stability and function of both Aβ fibrils and the adjacent cellular membrane. Taking advantage of the symmetry-related and directional properties of the protofilaments, we build models that cover several relative protofilament–membrane orientations. Our molecular dynamics simulations reveal the relative contributions of different structural elements to the dynamics and stability of Aβ protofilament segments near membranes, and the first steps in the mechanism of fibril–membrane interactions. During this process, we observe a significant alteration of the side-chain contact pattern in protofilaments, although a fraction of the characteristic β-sheet content is preserved. As a major driving force, we identify the electrostatic interactions between Aβ charged side chains, including E22, D23, and K28, and lipid headgroups. Together with hydrogen bonding with atoms from lipid headgroups, these interactions can facilitate the penetration of hydrophobic C-terminal amino acids through the lipid headgroup region, which can finally lead both to further loss of the initial fibril structure and to local membrane-thinning effects. Our results may guide new experiments that could test the extent to which the structural features of water-formed amyloid fibrils are preserved, lost, or reshaped by membrane-mediated interactions. © 2012 Elsevier Ltd. All rights reserved.

Introduction

*Corresponding author. School of Physics, University College Dublin, Belfield, Dublin 4, Ireland. E-mail address: [email protected]. Abbreviations used: Aβ, amyloid β; AD, Alzheimer's disease; MD, molecular dynamics; POPE, 1-palmitoyl-2oleoyl-sn-glycero-3-phosphoethanolamine; COM, center(s) of mass.

Alzheimer's disease (AD) amyloid β (Aβ) peptides, usually consisting of 38–43 amino acids, are cleavage products of β-secretases and γ-secretases acting on the transmembrane amyloid precursor protein. 1–3 Aβ peptides are the main component of extracellular depositions, which are one of the hallmarks of AD. 4,5 Upon cleavage, Aβ monomers are released into the extracellular environment, where they assemble into oligomers, fibrils, and, eventually, into filaments, which are the major

0022-2836/$ - see front matter © 2012 Elsevier Ltd. All rights reserved.

Aβ Protofilament-Lipid Membrane Interactions

components of amyloid plaques. In solution, in vitro, and in vivo, Aβ has been shown to form fibrils and oligomers 5–11; in lipid bilayers, experimental studies showed that introducing Aβ could lead to the formation of ion channels. 12–14 Amyloid toxicity is still not fully understood, although several mechanisms have been proposed to explain it. There have been several reports showing that Aβ is neurotoxic 15 and that its aggregation is required for toxicity. 16,17 Building on these, the “amyloid hypothesis”—stating that AD is the effect of Aβ accumulation and deposition—emerged, 18 while the “cascade hypothesis” regarded this accumulation as a most likely early event in the disease. 19 Although the Aβ fibrils found in amyloid plaques had been considered to be the main pathological markers in AD, 20 several recent studies have reported that, rather than the fully formed fibrils, protofibrils and soluble oligomers are the more toxic species. 4,5,21–25 These findings suggest that the end products of peptide aggregation in AD plaques (i.e., fibrils and high-order oligomers) may not be directly involved in cytotoxicity, as they are absent in certain forms of the disease and may not impair cell survival 26 but could actually favor it. 27 Prefibrillar oligomers have been reported to permeate cell membranes, while fibrils may lack this activity. 28,29 Nevertheless, calculations of the lateral accessible surface area of Aβ fibrils suggest that some polymorphic models have larger hydrophobic accessible surface area fractions than most proteins. 30 Prefibrillar oligomers can also bind to membranes and may transform into annular protofibrils that can form β-barrel membrane pores. 31 It has been suggested that the difference in toxicity between oligomers and fibrils may be due to the more exposed hydrophobic surfaces of oligomer β-sheets, which would be otherwise hidden by stacking interactions between the protofilaments in the mature fibrils. 32 Another possibility would be that the ends of the fibrillar structures are more reactive and, therefore, the toxic effects are weaker for fibrils than for shorter oligomers. 33 The smaller fibrillar oligomers may also be more toxic due to their reduced size and greater diffusion rates throughout the tissues, while fibrils could be less dangerous as they sequestrate the small oligomers in insoluble deposits. 23,34 Interestingly, observations on amyloid pathogenesis suggest a mechanism that involves the interactions of Aβ aggregates with cellular components that need to be accessible from both extracellular and cytosolic compartments, highlighting cell membranes as primary candidates. 35 Although the precise molecular mechanism of membrane permeation is unknown, several experiments using lipid bilayers demonstrate that amyloid oligomers interact with membranes and cause strong dysregulation of ion homeostasis. 33 Several studies have supported the idea that interactions between neuronal membranes and pep-

573 tides are a key element in the aggregation of the latter, leading to their cytotoxicity. 36–38 There are several pioneering experimental and computational studies relating the aggregation properties to the conformations adopted by Aβ peptides upon their interaction with lipids. 39–52 The aggregation process depends on the type of lipids (i.e., neutral or charged), peptide and salt concentrations, and pH. It has been reported that charged lipids induce β-strand structure in Aβ peptides to a similar extent as the increase in their concentration in solution, and that CD spectroscopy measurements can evidence significant (i.e., about 40–60%) β-strand content. 41 The presence of negatively charged lipids was similarly reported to be critical for the induction and stabilization of β-sheet structure, which in return leads to membrane perturbation. 42 Membrane permeation has emerged as a fundamental and widely-reported property of amyloid oligomers. The presence of Aβ oligomers seems to alter membrane structure and function and is related to nonspecific ion leakage. 29,53,54 Other studies support the idea of ion-specific channels. 12–14 The latter observation is supported by atomic force microscopy studies suggesting that Aβ peptides inserted into the lipid bilayer adopt an ion channel-like structure fragmented into several units surrounding a central pore-like region. 13,55,56 Several computational studies have already proposed and analyzed possible models of Aβ9–40 and Aβ17–42 preformed ion channels inserted into dioleoyl-glycero-phosphocholine and anionic lipid bilayers. 48,49,57 Using 20 ns-long to 50 ns-long molecular dynamics (MD) simulations, they reported that the initially continuous structures were subsequently fragmented into several oligomeric units that were loosely connected in the dioleoyl-glycero-phosphocholine bilayer to form transmembrane channels. Remarkably, these channel structures presented very similar numbers of oligomeric units and overall dimensions as the channels evidenced in previous atomic force microscopy experiments. 48,49,55–57 Studies supporting nonspecific ion leakage showed that the presence of amyloid oligomers can increase the conductance of the lipid bilayer by thinning the membrane and by increasing the area per lipid. 58 The effect of the oligomers on membrane conductance has been investigated by using single-ion channel detection techniques, although these studies did not evidence a significant formation of discrete ion channels by Aβ oligomers. 58 It was later noted that the observed different mechanisms of membrane permeation might be due to different experimental working conditions. 59 Since early studies of Aβ aggregation noted that different preparations of the peptides exhibited different aggregation kinetics, 60 and since recent studies evidence a significant structural polymorphism of Aβ fibrils, 30,61 it appears that there may be several membrane permeation mechanisms that are not mutually exclusive.

574 In this work, we investigate the early events that occur during the interactions between a preformed Aβ fibrillar oligomer (protofilament, ∼ 40 kDa) and a model bilayer membrane consisting of 1-palmitoyl2-oleoyl-sn-glycero-3-phosphoethanolamine (POPE) lipids. Our initial protofilament structures are built based on structural studies of Aβ protofilaments in solution 7,30,62 , 63—previously equilibrated both as parts of infinite fibrils and as finite fibril segments surrounded by water. 30 These fibril models have been studied in detail both under physiological conditions and elevated temperatures. 62 Here, we aim to analyze in atomistic detail the role of different structural elements in the stability and conformational dynamics of Aβ protofilaments near lipid bilayers. Our results may guide future studies of a similar system using coarse-grained models as well. 64,65 This study may shed new light on early molecular events that occur in processes such as membrane-mediated fibril nucleation or Aβ channel formation that are known to cause significant perturbations of normal physiological cell functions, as observed in AD and other amyloid-related diseases.

Results and Discussion In this study, we investigate at atomic-level detail the interactions between models of preformed Aβ9– 40 molecular protofilaments and lipid bilayer membranes in the presence of explicit water molecules. We emphasize from the start that the systems that we study are all molecular models, which, while simple and idealized, capture to the best of our knowledge the essential structural and dynamic features that have been revealed so far by previous experiments and computational modeling. 7,9,30,61,62,63,66 Studying these models offers unique insights into the detailed atomic-level mechanisms and processes involved in the early stages of interaction between Aβ protofilaments and lipid membranes. As in previous experimental and theoretical studies of structured peptide aggregates, the term “fibril” is used here for molecular structures made of aggregated peptides that have a rod-like appearance. 7,9,30,32,62 While fibrils can typically have significant β-sheet content, their exact degrees of molecular organization and structural order are generally low or unknown. Two or more amyloid fibrils can assemble laterally to form filaments. The term “protofilaments” is used for the basic structural units of structurally ordered amyloid filaments or fibrils and should not be confused with the term “protofibrils,” which is used generically for peptide aggregates that, while having a fibril-like structural appearance, are not yet fully formed fibrils. Here, we study Aβ ”protofilaments,” the basic structural units of Aβ fibrils consisting of “two-peptide”

Aβ Protofilament-Lipid Membrane Interactions

Fig. 1. Structural model of Aβ9–40 protofilaments (cross section) with a C2z topology. The main structural elements are as follows: N-terminal β-strands (red backbone), turn regions (green backbone) stabilized by D23-K28 salt bridges, and C-terminal β-strands (blue backbone). Charged side chains are shown in blue (positive) and red (negative). Other structurally important residues are also highlighted and labeled.

(dimeric) units formed by two Aβ9–40 peptides located in a plane roughly perpendicular to the fibril axis (Fig. 1). The two Aβ peptides in each twopeptide unit interact only through interdigitated side-chain contacts of their C-terminal β-sheets. As shown in Fig. 1, the Aβ40 models considered here correspond to protofilaments with a C 2z symmetry (where the z-direction is along the fibril axis), as these are the models that have so far been characterized most reliably in both previous experiments and simulations. 7,9,30,61,62,63,67 The N-terminal β-strand (red backbone), the turn region (green backbone) stabilized by the D23-K28 salt bridge, and the C-terminal β-strand (blue backbone) are some of the main structural features illustrated in Fig. 1. Charged side chains are shown in blue (positive) and red (negative), together with other structurally important residues that are also highlighted and labeled. These models have M35 residues in the Cterminal β-sheets forming interpeptide contacts in the core of the protofilament—a feature that plays an important role in fibril stability studies. 68 Interestingly, the intrinsic directionality of peptide backbone leads to specific directional asymmetries in the overall structural features of C2z Aβ protofilaments: the detailed chemical bonding properties at the two ends of the fibrils can be different. 62 As illustrated in Fig. 2, we can use a “right-hand rule” to discriminate between the “positive” (or “forward”) direction and the “negative” (“backward”) direction along the fibril axis. This aspect is potentially important, as future experiments that could measure disparities between the elongation and dissociation rates at fibril ends may use this information to infer

575

Aβ Protofilament-Lipid Membrane Interactions

Fig. 2. Top: Schematic representations of the initial conditions of the four systems studied (S1–S4), with different relative protofilament–membrane orientations. Only one of the four two-peptide layers is shown to illustrate fibril directionality. For S1 and S2, the fibril axis is perpendicular to the membrane, while for S3 and S4, it is parallel with the membrane surface. Bottom: More detailed molecular representations of the same initial conditions showing all four layers of the Aβ9–40 protofilament segments. For clarity, the explicit water molecules around the fibril and the membrane are not shown.

atomistic details about the molecular arrangements of the Aβ peptides in fibrils. Note that dimeric protofilaments with parallel Cterminal β-sheets (e.g., with C2x symmetry axes 30) would have both “+” and “−” regions at each end, thus presenting similar elongation/dissociation abilities at both ends. As the systems studied here have an anti-parallel C-terminal β-strand (i.e., C2z symmetry; see Fig. 2), they may have different elongation/dissociation properties at each fibril end, as well as different specific interactions with the membranes (e.g., see Fig. 6 and the associated discussion below). For our models, we note that there are four main possible relative orientations between protofilaments and the membrane, as illustrated in Fig. 2 for systems denoted by S1 and S2 (when the fibril axis is perpendicular to the membrane) and S3 and S4 (i.e., the fibril axis is parallel with the membrane). In S1, the “+” fibril axis points toward the membrane bilayer, while in S2, it is oriented away from it. In S3type systems, the N-terminal β-sheet is parallel with the membrane, while in S4-type systems, it is roughly normal to it. In the S4 case, the turn and N-terminal regions of the Aβ peptides participate in interactions at the interface region between the membrane and the fibrils. As described in Methods, we generate, equilibrate, and run long (i.e., 150 ns for each system) atomistic MD simulations, starting from the initial conditions that correspond to all these four cases (i.e., S1–S4, illustrated in the lower part of Fig. 2). Note that only the peptides and the model lipid membrane are

illustrated in the figure; water molecules are not shown here for clarity, although they are included explicitly in our simulations. Atomistic models of Aβ protofilament–lipid interactions The root mean square deviation (RMSD) for the Cα atoms, calculated over the four 150-ns trajectories, is illustrated in Fig. 3. Values calculated with respect to both the starting coordinates (RMSD0; red line) and the average coordinates (RMSDave; black line) respectively are shown. These RMSD curves indicate a significant loss of the initial structure of the Aβ protofilaments during the first steps of their interactions with the membranes before reaching an equilibrium-like conformational state. These RMSD0 values are significantly different, falling within the 8– 16 Å range from the initial structures corresponding to the equilibrated finite fibrils simulated extensively in previous studies in aqueous solutions. 30 To investigate these structural changes in more detail, we also calculate the RMSD with respect to the average structure along the same trajectories RMSDave after aligning to the initial frame. Using the frames with the lowest RMSDave (black line) in Fig. 3, we extract the representative structures for each case (see Fig. 3, side figures). In all cases, we observe a clear interaction of the parts of the protofilament with the membrane, especially for residues in the N-terminal β-strands (red) and the turn (green). This interaction is initiated mainly due to the presence of six charged residues in this region

576

Aβ Protofilament-Lipid Membrane Interactions

Fig. 3. Backbone Cα RMSD values calculated along the MD trajectories for the entire protofilament segments. Both values calculated with respect to the starting structure (RMSD0; red line) and the average structure (RMSDave; black line) are shown. Representative molecular structures corresponding to the smallest RMSDave value (circled) for each trajectory are shown in panels (a)-(d) next to each RMSD plot.

that are subjected to strong electrostatic interactions with the charged groups PO4− and NH3+ from the headgroups of the lipid molecules. We note that this initial electrostatic attraction is followed by the favorable interaction between the C-terminal region (blue; contains mainly hydrophobic amino acids and only a few polar amino acids) and the hydrophobic lipid tails. Secondary structure Secondary structure content was assessed using STRIDE 69 and monitored along the trajectories for all four cases. The predominant secondary structure elements—and a trademark of amyloid fibrils—are the N-terminal and C-terminal β-sheets. The βsheets run parallel with the fibril axis and are separated by distances normal to the fibril axis and about twice longer than typical hydrogen bonds, as evidenced by the typical cross-β pattern observed in X-ray fiber diffraction experiments and calculations. 9,30 NMR experiments indicate that Aβ1–40 fibrils are organized in hairpin-like structures, with the β-strands connected by a loop, which is further stabilized by a salt bridge. 7,30 The β-strands are formed in the region of residues 10–22 and 30–40, while residues 23–29 form a connecting loop. Solid-state NMR and X-ray crystallography experiments on amyloid-forming peptides show that the core of an amyloid fibril contains two or more layers of β-sheets. 7 In our simulations, the total content of the β-sheet decreases in each of the four systems. Figure 4a shows the fractions (%) of the α-helix and the βsheet for the whole fibril—the two main secondary structure elements observed during simulations. The structures that are not classified as α-helix or β-sheet are denoted generically as coil. For all four

systems, one can observe the obvious decrease in βsheet content to 40%, illustrating that individual monomers lose their initial β-rich conformations and degree of order in the fibrils, although not completely. At the same time, the content of the βsheet reaches a plateau region, which again indicates that the fibril loses its initial structure but is also stabilized overall by its new interaction with the membrane. There were very few helical conformations observed, mainly confined to one of the outer layers, the farthest from the membrane. The residues involved in these transitory α-helical structures in the turn region are S26NKGA30. For these outer peptides exposed to bulk water, the D23-K28 salt bridge is broken as the helix is formed, and the D23 and E22 residues compete for saltbridge interaction with K28. Due to the position of the three residues at the beginning and at the end of the helical turns, respectively, the position of E22 is generally more favorable for salt-bridge interactions with K28. 62,70 Assessing the secondary structure for each of the four layers in the fibril leads us to consider the two interior layers as forming a more stable “core” of our finite protofilament segments. Regarding the secondary structure of each of the four twopeptide layers in the fibril, this core region maintains a higher β-sheet content as compared to the peptides near the ends, which are exposed more strongly to interactions with either water molecules or lipid headgroups. These observations lead us to infer that (i) mainly the structure of the proximal layer in the fibril is affected by direct interactions with the membrane and that (ii) this loss in structure is not necessarily transmitted throughout the fibril to two-peptide units more distant from the membrane. 69

577

Aβ Protofilament-Lipid Membrane Interactions

on the nature of the hydrogen bond donor 71; in our analysis, we used a bond angle cutoff of 60° and an acceptor–donor distance of 3.1 Å. For systems S1 and S2, we have also used the analysis of hydrogen bonds (NHB) as an indicator of the interaction between the two-peptide layer proximal to the membrane and the lipid headgroups (layer 1 lipids, black line in Fig. 4b and Fig. SD2), as the phosphate and amino groups could serve as donors and acceptors for potential hydrogen bonds. Comparing it to the NHB between the dimeric layer closest to the membrane and the next layer in the fibril (layer 1–layer 2, red line in Fig. 4b and Fig. SD2), we found an anti-correlation between their values. We could thus attribute the loss of structure in layer 1 to the interactions with the lipid headgroups. C-terminal contacts

Fig. 4. Structural dynamics during protofilament– membrane interaction. The corresponding data for all four systems are reported in Figs. SD1, SD2, and SD3. (a) The secondary structure content along the MD trajectories. The β-strand fraction (black) decreases, while the coil fraction (red) increases. Only minor transient helix (blue) formation is observed. (b) Number of hydrogen bonds (NHB) between the two-peptide layers closest to the membrane (red) and between the end layer of the protofilament and the lipids (black). (c) The number of contacts between C-terminal β-strands decreases when the protofilament interacts more closely with the membrane.

Hydrogen bonds The hydrogen bonds between layers are directly responsible for the tertiary structural features of amyloid fibrils. 9 A loss in the number of backbone hydrogen bonds can be understood as a loss of the overall structural stability of the protofilament. As shown in Fig. 4b (see also Fig. SD2, Supplementary Data), our simulations reveal a clear initial decrease in the number of hydrogen bonds between the protofilament layers for all cases, after which the number of hydrogen bonds (NHB) remains relatively stable for the remainder of the 150-ns trajectory simulated for each case. This observation is correlated to the time evolution of the secondary structure pattern (see Fig. SD1). The geometric parameters for assessing hydrogen bonds depend

The protofilament orientation is characterized by contacts among amino acid side chains from adjacent β-sheets. Multiple contacts are possible, as fibrils grow even when the L17, L34, M35, and V36 were mutated to cysteine and L17C/L34C, L17C/ M35C, and L17C/V36C disulfidic bridges were formed. 9 In Fig. 4, one can clearly observe the decreasing value of C-terminal contacts between any of the amino acids in the A30-V40 region. Although the overall β-sheet content remains around 40% during the simulations, indicating that the β structure of individual monomers is relatively preserved, the low values for contacts between the C-termini reveal that the structure of the entire protofilaments is significantly altered. Interactions of charged residues and C-termini with lipids To illustrate how different parts of the protofilaments interact with the lipid headgroups or tails, Fig. 5 shows the projection, on the z-axis (approximately normal to the membrane), of the center(s) of mass (COM) of the charged amino acids (i.e., E11, H13, K16, E22, D23, and K28) situated in the N-terminus and turn regions (red line) to the COM of the P atoms in the lipid headgroups. The two red lines correspond to each of the two monomers in the dimeric layer proximal to the membrane for the S2 system. Also shown in Fig. 5 are the corresponding distances (blue lines) between the COM of the C-terminal β-strands and the same COM of the P atoms in lipids. The lipid headgroup region (pink) was approximated as the difference between the z-coordinates of the COM of N atoms and the z-coordinates of the COM of P atoms in the lipid headgroups. Our simulations have shown that, while the charged amino acids stay inside or slightly above the headgroup region

578

Aβ Protofilament-Lipid Membrane Interactions

Fig. 5. Protofilament–membrane interactions: the roles of the largely hydrophobic C-terminal β-strands and of the charged side chains for S2. The corresponding data for both S1 and S2 are reported in Fig. SD4. The plot shows projections, on the z-axis (i.e., approximately normal to the membrane), of the COM positions of the C-terminal β-strands (blue line) and of the charged residues in the N-terminal and turn regions (red line). The relative distances are calculated with respect to the position of the plane of the P atoms in the adjacent lipid layer. The pink area approximates the width of the lipids' headgroup region, calculated as the relative difference between the best-fit N-atom plane and the P-atom plane along the trajectory.

(red line), the C-termini can penetrate the headgroup region and interact directly with the lipid tails (blue line). 72 MD studies of monomeric Aβ in a membrane-like environment showed that the peptide is localized at the interface between the membrane and the solvent, and that it is likely to adopt helical structures. 73,74 Residues 1–28 were found to form electrostatic interactions with the polar lipid headgroups, causing the peptides to leave the lipid bilayer and to become localized atop the hydrophobic tail region. Thus, the preferential position in our trajectories of the charged residues above the membrane–solvent interface region was to be expected, as shown also by both experimental and computational studies involving monomeric Aβ on lipid membranes. 39,40,42,75,76,77 Figure 5 (and the corresponding Fig. SD4 in Supplementary Data) supports the idea that the strong initial electrostatic attraction between polar and charged Aβ side chains and lipid headgroups can be followed by direct interactions of the hydrophobic C-terminus with the lipid tails. Crossing this barrier may not be possible without the initial favorable electrostatic interaction. The predominant interactions are electrostatic; therefore, an increase in the salt concentration will screen the charges on both the peptide and the lipid heads, thus disrupting these electrostatic interactions and preventing the association of the peptide with negatively charged lipids, potentially reducing Aβ-related toxic effects on cellular membranes. 40,42 As seen in Fig. 5, although the C-terminus spends a significant part of the simulation time inserted in the lipid tail region during the 150-ns duration of our simulation, it does not adopt a helical structure, as suggested by experiments involving monomeric Aβ. The transient helix seen in the secondary structure plots belongs to one of the monomers in the two-peptide protofilament layer farthest from the membrane. At longer timescales, however, a membrane-inserted Aβ monomer may adopt a

significantly more helical structure. In our case, the peptide layer proximal to the membrane maintains some interactions with the layer above and thus acts differently from an isolated monomer. The effect of fibril on the membrane As a measure of the structural changes occurring in the membrane due to the interaction with the Aβ protofilament, we analyzed the relative orientations that the lipid tails adopted throughout the simulations. 14,37,78 To capture the effect of protofilaments on the membrane's structural order, we define the α angles between the lipid tails and the z-axis by using the relative positions of the lipids' P atoms and terminal C atoms (i.e., C218 and C316) with respect to the z-axis, as illustrated in Fig. 6a. Monitoring probabilities P(α) along our MD runs reveals a clear tendency of the C-terminal β-strands to interact more closely with the hydrophobic lipid tails for the S2 system (Fig. 6c) as compared to the S1 system (Fig. 6b). For S2, the C-termini that interact with the lipid tails attract them closer to the membrane's surface (i.e., toward large α angles), inducing a membrane-thinning effect. While some of the quantitative aspects of this tendency (as illustrated in Fig. 6b and c) may be force-fielddependent, this observation nevertheless points out differences in the detailed molecular mechanism of fibril membrane interactions between S1 and S2 cases. As mentioned above and as illustrated in Fig. 2, the detailed chemical differences at the ends of Aβ protofilaments could lead not only to asymmetric elongation kinetics but also to asymmetric interactions with lipid membranes. Internal hydration of Aβ fibrils An interesting aspect revealed by previous simulations 30,62 is the occurrence of narrow water channels—just a few (one to three) molecules wide— internal to the protofilaments and oriented along the

Aβ Protofilament-Lipid Membrane Interactions

579

Fig. 6. Effect of protofilaments on the membrane's structural order. (a) Definitions of α angles between the lipid tails and the z-axis. (b and c) Probabilities P(α) showing the tendency of the hydrophobic lipid tails to interact more closely with the C-terminal β-strands (i.e., large α values) for the S2 system (c) as compared to the S1 system (b).

fibril axis. Recent two-dimensional infrared spectroscopy experiments have evidenced the possibility that Aβ40 fibrils indeed contain structurally significant mobile water molecules within the intersheet region. 79 Several subsequent simulations and experiments support this finding as well. 80–84 In the simulations of this study, because we use the preequilibrated Aβ9–40 protofilaments as initial structures, 30,62 the water molecules are found to hydrate internal side chains from the beginning in all four systems. As the interaction between the model fibrils and the lipids progresses, the structures of the protofilaments become more open (e.g., see Fig. 3), and many of the previously confined water molecules become more exposed to the bulk water. An interesting aspect, discussed in more detail in Salt Bridges and Electrostatic Interactions, is that the water molecules hydrating significantly polar amino acids, in particular the D23-K28 salt bridges, become involved in mediating their subsequent interactions with other charged residues and, more importantly, with the lipid headgroups. Salt bridges and electrostatic interactions The salt bridge between D23 and K28 in the loop region is an important structural element, stabilizing the bend segment formed by residues in the D23-V29 region. Its importance has been demonstrated by numerous experimental and computational studies 7,63,67 and particularly by Sciarretta et al. 85 Their solid-state NMR study measured the aggregation rates for Aβ40 fibrils synthesized with an amide

cross-link between D23 and K28 (Aβ1–40 lactam(D23/ K28)). The results showed that the bend-forming step is rate-limiting, inducing a lag period in the fibril formation; thus, peptides with an already formed salt bridge could surpass this stage and form fibrils much faster. Although the wild type and the crosslinked fibril types have similar structures, Aβ1–40 lactam(D23/K28) aggregates at a rate 3 orders of magnitude higher than the wild type. 85 A subsequent computational study by Reddy et al. further relates the rate enhancement to the decrease in free-energy barrier (estimated in the range of 4–7 kBT), which correlates to the energy cost in forming the bend. 86 In the study of Buchete et al., 30 simulations of finite-size Aβ9–40 protofilaments showed that these maintain an ordered structure in bulk water (in agreement with experiments), 7 being stabilized by both intramolecular and intermolecular D23-K28 staggered salt bridges, partially hydrated by narrow water channels oriented almost parallel with the fibril axis (see Internal Hydration of Aβ Fibrils). In this study, we estimate the stability of the D23K28 salt bridges by measuring the electrostatic energies between the negatively-charged D23 residues and the positively-charged K28 residues. There is rapid growth in the energies for D23-K28, showing how quickly the salt bridge loses stability (Fig. 7c, red line), with values starting from about − 200 kcal/ mol reaching and converging toward 0 kcal/mol (i.e., no interaction). The black line shows the interaction between the salt-bridge residues in the monomers closest to the membrane and the lipids having lower but fluctuating values. This is due to

580

Aβ Protofilament-Lipid Membrane Interactions

Fig. 7. Role of the D23-K28 salt bridge in protofilament–membrane interactions. (a) Illustration of an initially intact D23-K28 salt bridge for an Aβ monomer. (b) During the simulation, when in close proximity to the membrane, the D23 and K28 side chains interact with the charged lipid headgroup. (c) Electrostatic energy for the D23-K28 salt bridge in the first two-peptide layer (red) and for the interaction between the D23 and K28 side chains of the first layer and the lipids (blue). (d) The electrostatic energies between E22-K28 (black) and D23-K28 (red), plotted for the entire fibril.

the dynamic environment formed by the fibrils, water, and lipids, where the fibrils are not trapped by a subset of lipids but span a wide array of lipid molecules. We also compared the electrostatic energy for the D23-K28 pair (see Fig. 7), and we observed that the salt bridges in the core of the protofilament (i.e., the two middle peptide layers) are more stable than the salt bridges at the ends due to interaction with either bulk water or lipids. The two salt bridges for the layer closest to the membrane are the least stable due to interaction with the membrane, as illustrated for example in Fig. 7b. The second least stable components are the salt bridges in the layer exposed to bulk water, distal to the membrane. The core of the protofilament, formed by the two inner layers, maintains significantly more stable D23-K28 salt bridges, as it is more protected from the interaction with either lipids or water. Moreover, in the case of the peptide layer closest to the membrane when the two residues lose contact with each other, they interact with the lipid headgroups (i.e., the negatively charged aspartic acid D23 interacts with the positively charged amino group of the lipids, while the lysine K28 interacts with the negatively charged phosphate groups), as depicted in Fig. 7b. The interaction with the headgroups can be observed for all the charged residues of the first peptide layer of the protofilament, as they tend to reorient themselves toward the lipids. To understand the complex role of D23-K28 interactions with the lipid headgroups, one should also account for the possible competition between negatively charged partners E22 and D23 for the positively charged K28. Previous simulations in bulk water showed that this could be a destabilizing factor for maintaining the β-strand structure and can be particularly important during the association/dissociation kinetics of the monomers at the fibril ends. 62,87,88 Here, in the presence of lipids, the

probability of competitive E22-K28 salt-bridge formation is still observed (although it is greatly reduced), as illustrated by the corresponding electrostatic energies shown in Fig. 7d for S1 (and in Fig. SD6 for all four systems). While still present, this competition is strongly affected by the more favorable interactions between the peptide side chains and the charged lipid headgroups, as illustrated in Fig. 7b.

Conclusions In this work, we explored in atomic-level detail the early steps of the molecular interactions between models of lipid membranes and Aβ40 protofilaments, using MD and explicit water molecules. Our approach is enabled by previous experimental and computational studies 7,30,62 that described the structural properties and conformational dynamics of possible models of Aβ40 protofilaments. By taking advantage of the local symmetry-related and directional properties of the protofilament segments, we build models that cover a broad range of relative protofilament–membrane orientations, corresponding to systems denoted by S1–S4 and illustrated in Fig. 2. These systems were subsequently simulated in the NPT ensemble for more than 150 ns in each case (see Methods and Table 1). While some of our observations may be force field-specific and model-specific, they nevertheless suggest some general principles that control the first events during the interactions of amyloid protofilaments with lipid bilayer membranes. For example, we find that, in all cases, the fibrils adopt new structures upon interaction with the lipids, quite different from the initial conformations, as demonstrated by the increase in RMSD0 values shown in Fig. 3. As this increase is larger than that in the case of systems in aqueous solution, 30,62 we attribute it to

581

Aβ Protofilament-Lipid Membrane Interactions

Table 1. MD parameters of the systems of Aβ protofilaments interacting with lipid bilayer membranes and with explicit TIP3P water molecules System components Aβ protofilament System label Peptides (Np) Atoms S1

2×4=8

3824

S2

2×4=8

3824

S3

2×4=8

3824

S4

2×4=8

3824

Lipid bilayer

Orientation Fibril axis normal, away from membrane Fibril axis normal, toward membrane Fibril axis parallel, N-terminus near membrane Fibril axis parallel, turn near membrane

NA

System size

POPE lipid bilayer

Water Initial Atoms Molecules (Nl) Atoms molecules (Nw) dimensions (Å3) (Na) 2 × 7 × 7 = 98

12,250

8960

82 × 82 × 90

42,954

2 × 7 × 7 = 98

12,250

8831

82 × 82 × 90

42,567

2 × 7 × 7 = 98

12,250

10,193

82 × 82 × 100

46,653

2 × 7 × 7 = 98

12,250

13,810

71 × 71 × 126

57,504

2 × 7 × 7 = 98

12,250

11,800

82 × 82 × 115

47,650

NA, not applicable. Each of the four systems was subjected to 150 ns of MD simulations in the isothermal–isobaric ensemble.

the specific influence of peptide–lipid interactions. The fact that the RMSD values nevertheless reach a plateau suggests that our systems reach quasiequilibrated states, which could be subjected to further conformational changes on longer timescales. While different from the initial conditions, our final quasi-equilibrated conformational states of membrane-bound amyloid peptides preserve a large fraction of their characteristic β-sheet content, thus retaining the interpeptide hydrogen-bonding ability of the Aβ peptides in protofilaments. The βsheet content is particularly better preserved closer to the core of our protofilaments (i.e., in the inner two-peptide layers). The layers at the end of the fibril have more significantly perturbed structures due to the interaction with either the lipids or the bulk water. By analyzing the number of hydrogen bonds between the first two layers in the protofilament proximal to the membrane (i.e., layers 1 and 2) and between layer 1 and lipids, we found an anticorrelation that proves that the loss of structure in the fibril is due to specific interactions with the lipids. This conclusion is further supported by the analysis of the electrostatic energies of interaction between the charged residues (such as D23 and K28) and the lipid headgroups. The case of D23 and K28 residues is particularly relevant, as they are involved in stabilizing salt bridges present in Aβ protofilaments formed in solution. 7,30,62,87 In the presence of the membrane, we observe a strong destabilization of these salt bridges due to competing favorable electrostatic interactions with the lipid headgroups. Interestingly, even the competing E22K28 salt bridges, while also present, have relatively weak interaction energies and, thus, stability as compared to the interactions with lipids. We investigate also the interactions between the more hydrophobic C-terminal β-strand regions and the lipid tails, and observe not only their occurrence but also their effect on the local order and orientation

of the lipid tails. During interactions with the protofilaments, we showed a significant bias for lipid tails to be oriented away from the core region, which could lead to local membrane-thinning effects. In summary, our study shows that the early steps in the interaction mechanism between Aβ protofilaments and lipid membranes are driven by the electrostatic interactions between several charged peptide side chains and the lipid headgroups. These interactions can partially destabilize the β–turn–β motif both by perturbing the D23-K28 salt bridges and by bringing the peptides closer overall to the membrane. Together with the possibility of hydrogen bonding with atoms from the lipid headgroups, this leads to an initial loss of β-sheet content within the protofilaments. Subsequently, the hydrophobic Cterminal amino acids may penetrate the lipid headgroup region and interact directly with the hydrophobic lipid tails. This can lead to further loss of the initial fibril structure and local membrane-thinning effects. Although the times accessible to our atomistic MD simulations are much shorter than processes related to possible amyloid pore formation, our simulations show that this type of unmediated interaction with lipid membranes could lead to peptide structures that will lack a significant fraction of the rich β-sheet content and the stabilizing salt bridges characteristic of the β–turn–β motif of water-formed Aβ protofilaments. Alternatively, an unusually complex molecular mechanism may be required to protect the salt bridges and β-sheet regions of water-formed amyloids during their membrane formation or insertion. We hope that the observations reported in this study could lead to future experiments that could probe the local peptide structure in different environments (e.g., solid-state or liquid-state NMR 32,89,90 and two-dimensional infrared spectroscopy, 79,91–93 coupled with mutations of charged side chains or hydrogen/deuterium exchange, etc.), which may be able to test the extent to which the side-chain contacts

582 and the β-sheet content specific for water-formed Aβ fibrils are preserved, lost, or reshaped by membrane interactions. These considerations could advance the understanding of membrane-mediated fibril nucleation or Aβ channel formation processes that can cause significant perturbations of normal physiological cell membrane functions, as observed in AD and other amyloid-related diseases.

Methods Molecular dynamics A summary of the main MD parameters used in this study is given in Table 1 and presented in detail below. The initial conformations for the four systems illustrated in Fig. 2 were generated based on previously performed large-scale simulations of the Aβ9–40 protofilaments in bulk water. 30,62 The Visual Molecular Dynamics software 94 was used for modeling, data visualization, and analysis. All four systems (S1–S4) were built from an equilibrated model POPE lipid membrane, similarly to previous studies. 95,96 Each system of lipids, peptides, and water consisted of about 42,000–57,000 atoms, as reported in Table 1. The allatom MD simulations in this study were performed using the NAMD software 97 (with the CHARMM27 force field), 98 NPT ensemble (i.e., constant number of molecules, pressure, and temperature), periodic boundary conditions, and the same parameters as in our previous work. 30,62 The temperature was set to 310 K, close to physiological conditions. The pressure was maintained constant at 1 atm using the Langevin piston method, and the temperature was controlled using Langevin dynamics. 99,100 The peptide– lipid systems were solvated with explicit solvent using TIP3P water molecules. 101 The integration step was 1 fs. Before the simulations of the combined lipid–peptide systems could be performed, the POPE bilayers were first solvated, minimized, heated, and equilibrated. Subsequently, an NPT simulation of only the membrane in water was run in order to ensure that the lipid membrane was stable and preequilibrated. The lipid bilayer consisted of 98 POPE molecules. Each of the two membrane layers consists of 49 POPE lipids arranged approximately on a 7 × 7 initial grid, with an initial distance of 12 Å in the layer that changes subsequently during the preequilibration run. We considered the bilayer membrane to be stable during the preequilibration NPT simulation when the distance between the COM of the two layers converged to a relatively constant value of about 31 Å. As a supplementary check, we built membranes with the same number of lipids, but we minimized, heated, and equilibrated at different rates. The two resulting membranes had the distance between the COM converging to a similar value, providing the model membrane that we used in the ulterior MD simulations of the systems of peptide fibrils in contact with lipid membranes. Aβ protofilament–membrane simulations In vivo, Aβ1–40 has been shown to be the most prevalent type of peptide found in the cerebrospinal

Aβ Protofilament-Lipid Membrane Interactions fluid. 102,103 In our study, we simulated fibrils of the two-peptide layers of wild-type Aβ9–40, which is used to model the full Aβ1–40. The first eight residues in the N-terminus are significantly more disordered and were not considered in previous fibril models. 9,61 The sequence of the peptide is G9YEVHHQKLVFFAED23 VGSNK28GAIIGLMVGGVV40, with two β-strands in the V12-V24 and A30-V40 regions connected by a turn stabilized by a salt bridge between D23 and K28, as illustrated in Fig. 1. All fibrils in this study have the C2z topology 62 as it is most strongly supported by experiments. 9,63 The experimentally solved structures were first simulated in aqueous solution as an infinite fibril system 62 from which an octamer consisting of four two-peptide layers was extracted and simulated in the presence of the lipid membrane in this study. In constructing the initial conditions for our MD simulations, we took into consideration the relative fibril–membrane orientations (see Fig. 2), as indicated by the sign determined by the intrinsic fibril direction and as defined by a right-hand rule for the N-terminus-to-Cterminus direction of the peptide. 62 We constructed four systems (denoted S1–S4) with diverse conformations, placing the peptides on the lipid membrane model surrounded by TIP3P water molecules. The four conformations shown in Fig. 2 cover all the main possibilities of relative fibril–membrane orientations and account for various interaction modes between the residue side chains in the fibril and the lipids. As summarized in Table 1, the total simulation time was larger than 0.6 μs—not considering the preequilibration runs for lipids and the protofilaments presented in previous work. 30,62

Acknowledgements We thank Gerhard Hummer, John Straub, Devarajan Thirumalai, and Robert Tycko for helpful discussions. We gratefully acknowledge the Irish Research Council for Science, Engineering, and Technology for financial support and the Irish Centre for High-End Computing for the use of computational facilities.

Supplementary Data Supplementary data to this article can be found online at doi:10.1016/j.jmb.2011.12.063

References 1. Miller, D. L., Papayannopoulos, I. A., Styles, J., Bobin, S. A., Lin, Y. Y., Biemann, K. & Iqbal, K. (1993). Peptide compositions of the cerebrovascular and senile plaque core amyloid deposits of Alzheimer's disease. Arch. Biochem. Biophys. 301, 41–52.

583

Aβ Protofilament-Lipid Membrane Interactions 2. Vassar, R. & Citron, M. (2000). Abeta-generating enzymes: recent advances in beta- and gammasecretase research. Neuron, 27, 419–422. 3. Wolfe, M. S. & Guenette, S. Y. (2007). APP at a glance. J. Cell Sci. 120, 3157–3161. 4. Walsh, D. T., Montero, R. M., Bresciani, L. G., Jen, A. Y., Leclercq, P. D., Saunders, D. et al. (2002). Amyloid-beta peptide is toxic to neurons in vivo via indirect mechanisms. Neurobiol. Dis. 10, 20–27. 5. Hardy, J. & Selkoe, D. J. (2002). The amyloid hypothesis of Alzheimer's disease: progress and problems on the road to therapeutics. Science, 297, 353–356. 6. Serpell, L. C. (2000). Alzheimer's amyloid fibrils: structure and assembly. Biochim. Biophys. Acta, 1502, 16–30. 7. Petkova, A. T., Ishii, Y., Balbach, J. J., Antzutkin, O. N., Leapman, R. D., Delaglio, F. & Tycko, R. (2002). A structural model for Alzheimer's betaamyloid fibrils based on experimental constraints from solid state NMR. Proc. Natl Acad. Sci. USA, 99, 16742–16747. 8. Tycko, R. (2004). Progress towards a molecular-level structural understanding of amyloid fibrils. Curr. Opin. Struct. Biol. 14, 96–103. 9. Petkova, A. T., Yau, W. M. & Tycko, R. (2006). Experimental constraints on quaternary structure in Alzheimer's beta-amyloid fibrils. Biochemistry, 45, 498–512. 10. Walsh, D. M., Hartley, D. M., Kusumoto, Y., Fezoui, Y., Condron, M. M., Lomakin, A. et al. (1999). Amyloid beta-protein fibrillogenesis. Structure and biological activity of protofibrillar intermediates. J. Biol. Chem. 274, 25945–25952. 11. Luhrs, T., Ritter, C., Adrian, M., Riek-Loher, D., Bohrmann, B., Dobeli, H. et al. (2005). 3D structure of Alzheimer's amyloid-beta(1–42) fibrils. Proc. Natl Acad. Sci. USA, 102, 17342–17347. 12. Arispe, N., Pollard, H. B. & Rojas, E. (1994). Betaamyloid Ca(2+)-channel hypothesis for neuronal death in Alzheimer disease. Mol. Cell Biochem. 140, 119–125. 13. Lashuel, H. A., Hartley, D., Petre, B. M., Walz, T. & Lansbury, P. T., Jr (2002). Neurodegenerative disease: amyloid pores from pathogenic mutations. Nature, 418, 291. 14. Lau, T. L., Ambroggio, E. E., Tew, D. J., Cappai, R., Masters, C. L., Fidelio, G. D. et al. (2006). Amyloidbeta peptide disruption of lipid membranes and the effect of metal ions. J. Mol. Biol. 356, 759–770. 15. Yankner, B. A., Duffy, L. K. & Kirschner, D. A. (1990). Neurotrophic and neurotoxic effects of amyloid beta protein: reversal by tachykinin neuropeptides. Science, 250, 279–282. 16. Pike, C. J., Walencewicz, A. J., Glabe, C. G. & Cotman, C. W. (1991). In vitro aging of beta-amyloid protein causes peptide aggregation and neurotoxicity. Brain Res. 563, 311–314. 17. Busciglio, J., Lorenzo, A. & Yankner, B. A. (1992). Methodological variables in the assessment of beta amyloid neurotoxicity. Neurobiol. Aging, 13, 609–612. 18. Shankar, G. M., Li, S., Mehta, T. H., Garcia-Munoz, A., Shepardson, N. E., Smith, I. et al. (2008). Amyloid-beta protein dimers isolated directly from Alzheimer's

19. 20. 21.

22.

23.

24.

25.

26. 27.

28.

29.

30. 31.

32.

33. 34.

brains impair synaptic plasticity and memory. Nat. Med. 14, 837–842. Hardy, J. A. & Higgins, G. A. (1992). Alzheimer's disease: the amyloid cascade hypothesis. Science, 256, 184–185. Rochet, J. C. & Lansbury, P. T., Jr (2000). Amyloid fibrillogenesis: themes and variations. Curr. Opin. Struct. Biol. 10, 60–68. Haass, C. & Selkoe, D. J. (2007). Soluble protein oligomers in neurodegeneration: lessons from the Alzheimer's amyloid beta-peptide. Nat. Rev. Mol. Cell Biol. 8, 101–112. Gong, Y., Chang, L., Viola, K. L., Lacor, P. N., Lambert, M. P., Finch, C. E. et al. (2003). Alzheimer's disease-affected brain: presence of oligomeric A beta ligands (ADDLs) suggests a molecular basis for reversible memory loss. Proc. Natl Acad. Sci. USA, 100, 10417–10422. Lambert, M. P., Barlow, A. K., Chromy, B. A., Edwards, C., Freed, R., Liosatos, M. et al. (1998). Diffusible, nonfibrillar ligands derived from Abeta1– 42 are potent central nervous system neurotoxins. Proc. Natl Acad. Sci. USA, 95, 6448–6453. Hartley, D. M., Walsh, D. M., Ye, C. P., Diehl, T., Vasquez, S., Vassilev, P. M. et al. (1999). Protofibrillar intermediates of amyloid beta-protein induce acute electrophysiological changes and progressive neurotoxicity in cortical neurons. J. Neurosci. 19, 8876–8884. Kirkitadze, M. D., Bitan, G. & Teplow, D. B. (2002). Paradigm shifts in Alzheimer's disease and other neurodegenerative disorders: the emerging role of oligomeric assemblies. J. Neurosci. Res. 69, 567–577. Tompkins, M. M. & Hill, W. D. (1997). Contribution of somal Lewy bodies to neuronal death. Brain Res. 775, 24–29. Arrasate, M., Mitra, S., Schweitzer, E. S., Segal, M. R. & Finkbeiner, S. (2004). Inclusion body formation reduces levels of mutant huntingtin and the risk of neuronal death. Nature, 431, 805–810. Janson, J., Ashley, R. H., Harrison, D., McIntyre, S. & Butler, P. C. (1999). The mechanism of islet amyloid polypeptide toxicity is membrane disruption by intermediate-sized toxic amyloid particles. Diabetes, 48, 491–498. Demuro, A., Mina, E., Kayed, R., Milton, S. C., Parker, I. & Glabe, C. G. (2005). Calcium dysregulation and membrane disruption as a ubiquitous neurotoxic mechanism of soluble amyloid oligomers. J. Biol. Chem. 280, 17294–17300. Buchete, N. V., Tycko, R. & Hummer, G. (2005). Molecular dynamics simulations of Alzheimer's betaamyloid protofilaments. J. Mol. Biol. 353, 804–821. Kayed, R., Pensalfini, A., Margol, L., Sokolov, Y., Sarsoza, F., Head, E. et al. (2009). Annular protofibrils are a structurally and functionally distinct type of amyloid oligomer. J. Biol. Chem. 284, 4230–4237. Paravastu, A. K., Leapman, R. D., Yau, W. M. & Tycko, R. (2008). Molecular structural basis for polymorphism in Alzheimer's beta-amyloid fibrils. Proc. Natl Acad. Sci. USA, 105, 18349–18354. Glabe, C. G. (2009). Amyloid oligomer structures and toxicity. Open Biol. J. 2, 222–227. Xue, W. F., Hellewell, A. L., Gosal, W. S., Homans, S. W., Hewitt, E. W. & Radford, S. E. (2009). Fibril

584

35.

36.

37.

38.

39. 40.

41.

42.

43. 44. 45. 46. 47.

48.

49.

50.

Aβ Protofilament-Lipid Membrane Interactions fragmentation enhances amyloid cytotoxicity. J. Biol. Chem. 284, 34272–34282. Kayed, R., Head, E., Thompson, J. L., McIntire, T. M., Milton, S. C., Cotman, C. W. & Glabe, C. G. (2003). Common structure of soluble amyloid oligomers implies common mechanism of pathogenesis. Science, 300, 486–489. Kremer, J. J., Pallitto, M. M., Sklansky, D. J. & Murphy, R. M. (2000). Correlation of beta-amyloid aggregate size and hydrophobicity with decreased bilayer fluidity of model membranes. Biochemistry, 39, 10309–10318. Mingeot-Leclercq, M. P., Lins, L., Bensliman, M., Van Bambeke, F., Van Der Smissen, P., Peuvot, J. et al. (2002). Membrane destabilization induced by betaamyloid peptide 29–42: importance of the aminoterminus. Chem. Phys. Lipids, 120, 57–74. Bokvist, M., Lindstrom, F., Watts, A. & Grobner, G. (2004). Two types of Alzheimer's beta-amyloid (1–40) peptide membrane interactions: aggregation preventing transmembrane anchoring versus accelerated surface fibril formation. J. Mol. Biol. 335, 1039–1049. Terzi, E., Holzemann, G. & Seelig, J. (1997). Interaction of Alzheimer beta-amyloid peptide(1–40) with lipid membranes. Biochemistry, 36, 14845–14852. Hertel, C., Terzi, E., Hauser, N., JakobRotne, R., Seelig, J. & Kemp, J. A. (1997). Inhibition of the electrostatic interaction between beta-amyloid peptide and membranes prevents beta-amyloid-induced toxicity. Proc. Natl Acad. Sci. USA, 94, 9412–9416. Terzi, E., Holzemann, G. & Seelig, J. (1995). Selfassociation of beta-amyloid peptide(1–40) in solution and binding to lipid-membranes. J. Mol. Biol. 252, 633–642. McLaurin, J. & Chakrabartty, A. (1997). Characterization of the interactions of Alzheimer beta-amyloid peptides with phospholipid membranes. Eur. J. Biochem. 245, 355–363. Yip, C. M., Darabie, A. A. & McLaurin, J. (2002). Abeta42-peptide assembly on lipid bilayers. J. Mol. Biol. 318, 97–107. Yip, C. M. & McLaurin, J. (2001). Amyloid-beta peptide assembly: a critical step in fibrillogenesis and membrane disruption. Biophys. J. 80, 1359–1371. McLaurin, J., Yang, D., Yip, C. M. & Fraser, P. E. (2000). Review: modulating factors in amyloid-beta fibril formation. J. Struct. Biol. 130, 259–270. Morriss-Andrews, A., Bellesia, G. & Shea, J. E. (2011). Effects of surface interactions on peptide aggregate morphology. J. Chem. Phys. 135, 085102. Martinez, A. V., DeSensi, S. C., Dominguez, L., Rivera, E. & Straub, J. E. (2011). Protein folding in a reverse micelle environment: the role of confinement and dehydration. J. Chem. Phys. 134, 055107. Jang, H., Arce, F. T., Ramachandran, S., Capone, R., Lal, R. & Nussinov, R. (2010). Beta-barrel topology of Alzheimer's beta-amyloid ion channels. J. Mol. Biol. 404, 917–934. Jang, H., Zheng, J. & Nussinov, R. (2007). Models of beta-amyloid ion channels in the membrane suggest that channel formation in the bilayer is a dynamic process. Biophys. J. 93, 1938–1949. Williams, T. L. & Serpell, L. C. (2011). Membrane and surface interactions of Alzheimer's A beta peptide—

51.

52.

53.

54.

55.

56.

57.

58.

59. 60.

61.

62.

63.

64.

65.

insights into the mechanism of cytotoxicity. FEBS J. 278, 3905–3917. Campioni, S., Mannini, B., Zampagni, M., Pensalfini, A., Parrini, C., Evangelisti, E. et al. (2010). A causative link between the structure of aberrant protein oligomers and their toxicity. Nat. Chem. Biol. 6, 140–147. Strodel, B., Lee, J. W. L., Whittleston, C. S. & Wales, D. J. (2010). Transmembrane structures for Alzheimer's A beta(1–42) oligomers. J. Am. Chem. Soc. 132, 13300–13312. Kayed, R., Sokolov, Y., Edmonds, B., McIntire, T. M., Milton, S. C., Hall, J. E. & Glabe, C. G. (2004). Permeabilization of lipid bilayers is a common conformation-dependent activity of soluble amyloid oligomers in protein misfolding diseases. J. Biol. Chem. 279, 46363–46366. Sokolov, Y. V., Kayed, R., Kozak, A., Edmonds, B., McIntire, T. M., Milton, S. et al. (2004). Soluble amyloid oligomers increase lipid bilayer conductance by increasing the dielectric constant of the hydrocarbon core. Biophys. J. 86, 382a. Lin, H., Bhatia, R. & Lal, R. (2001). Amyloid beta protein forms ion channels: implications for Alzheimer's disease pathophysiology. FASEB J. 15, 2433–2444. Lal, R., Lin, H. & Quist, A. P. (2007). Amyloid beta ion channel: 3D structure and relevance to amyloid channel paradigm. Biochim. Biophys. Acta, 1768, 1966–1975. Jang, H., Arce, F. T., Capone, R., Ramachandran, S., Lal, R. & Nussinov, R. (2009). Misfolded amyloid ion channels present mobile beta-sheet subunits in contrast to conventional ion channels. Biophys. J. 97, 3029–3037. Sokolov, Y., Kozak, J. A., Kayed, R., Chanturiya, A., Glabe, C. & Hall, J. E. (2006). Soluble amyloid oligomers increase bilayer conductance by altering dielectric structure. J. Gen. Physiol. 128, 637–647. Eliezer, D. (2006). Amyloid ion channels: a porous argument or a thin excuse? J. Gen. Physiol. 128, 631–633. Soto, C., Castano, E. M., Kumar, R. A., Beavis, R. C. & Frangione, B. (1995). Fibrillogenesis of synthetic amyloid-beta peptides is dependent on their initial secondary structure. Neurosci. Lett. 200, 105–108. Paravastu, A. K., Petkova, A. T. & Tycko, R. (2006). Polymorphic fibril formation by residues 10–40 of the Alzheimer's beta-amyloid peptide. Biophys. J. 90, 4618–4629. Buchete, N. V. & Hummer, G. (2007). Structure and dynamics of parallel beta-sheets, hydrophobic core, and loops in Alzheimer's A beta fibrils. Biophys. J. 92, 3032–3039. Petkova, A. T., Leapman, R. D., Guo, Z., Yau, W. M., Mattson, M. P. & Tycko, R. (2005). Self-propagating, molecular-level polymorphism in Alzheimer's betaamyloid fibrils. Science, 307, 262–265. Buchete, N. V., Straub, J. E. & Thirumalai, D. (2008). Dissecting contact potentials for proteins: relative contributions of individual amino acids. Proteins Struct. Funct. Genet. 70, 119–130. Buchete, N. V., Straub, J. E. & Thirumalai, D. (2004). Orientation-dependent coarse-grained potentials

585

Aβ Protofilament-Lipid Membrane Interactions

66. 67.

68.

69. 70.

71.

72.

73.

74.

75. 76.

77.

78.

79.

80.

derived by statistical analysis of molecular structural databases. Polymer, 45, 597–608. Berezhkovskii, A. M., Tofoleanu, F. & Buchete, N. V. (2011). Are peptides good two-state folders? J. Chem. Theory Comput. 7, 2370–2375. Ma, B. & Nussinov, R. (2002). Stabilities and conformations of Alzheimer's beta-amyloid peptide oligomers (Abeta 16–22, Abeta 16–35, and Abeta 10–35): sequence effects. Proc. Natl Acad. Sci. USA, 99, 14126–14131. O'Brien, E. P., Okamoto, Y., Straub, J. E., Brooks, B. R. & Thirumalai, D. (2009). Thermodynamic perspective on the dock–lock growth mechanism of amyloid fibrils. J. Phys. Chem. B, 113, 14421–14430. Frishman, D. & Argos, P. (1995). Knowledge-based protein secondary structure assignment. Proteins Struct. Funct. Genet. 23, 566–579. Miyashita, N., Straub, J. E. & Thirumalai, D. (2009). Structures of beta-amyloid peptide 1–40, 1–42, and 1– 55—the 672–726 fragment of APP—in a membrane environment with implications for interactions with gamma-secretase. J. Am. Chem. Soc. 131, 17843–17852. Legon, A. C. & Millen, D. J. (1987). Angular geometries and other properties of hydrogen-bonded dimers—a simple electrostatic interpretation of the success of the electron-pair model. Chem. Soc. Rev. 16, 467–498. Demeester, N., Baier, G., Enzinger, C., Goethals, M., Vandekerckhove, J., Rosseneu, M. & Labeur, C. (2000). Apoptosis induced in neuronal cells by Cterminal amyloid beta-fragments is correlated with their aggregation properties in phospholipid membranes. Mol. Membr. Biol. 17, 219–228. Ma, J. P., Xu, Y. C., Shen, J. J., Luo, X. M., Zhu, W. L., Chen, K. X. & Jiang, H. L. (2005). Conformational transition of amyloid beta-peptide. Proc. Natl Acad. Sci. USA, 102, 5403–5407. Straub, J. E., Miyashita, N. & Thirumalai, D. (2009). Structures of beta-amyloid peptide 1–40, 1–42, and 1– 55—the 672–726 fragment of APP—in a membrane environment with implications for interactions with gamma-secretase. J. Am. Chem. Soc. 131, 17843–17852. McLaurin, J., Yip, C. M. & Darabie, A. A. (2002). A beta 42-peptide assembly on lipid bilayers. J. Mol. Biol. 318, 97–107. Chong, S. H. & Ham, S. (2011). Atomic decomposition of the protein solvation free energy and its application to amyloid-beta protein in water. J. Chem. Phys. 135, 034506. Lee, C. & Ham, S. (2011). Characterizing amyloidbeta protein misfolding from molecular dynamics simulations with explicit water. J. Comput. Chem. 32, 349–355. Pastor, R. W., Klauda, J. B., Roberts, M. F., Redfield, A. G. & Brooks, B. R. (2008). Rotation of lipids in membranes: molecular dynamics simulation, P-31 spin–lattice relaxation, and rigid-body dynamics. Biophys. J. 94, 3074–3083. Kim, Y. S., Liu, L., Axelsen, P. H. & Hochstrasser, R. M. (2009). 2D IR provides evidence for mobile water molecules in beta-amyloid fibrils. Proc. Natl Acad. Sci. USA, 106, 17751–17756. Miller, Y., Ma, B. Y., Tsai, C. J. & Nussinov, R. (2010). Hollow core of Alzheimer's A beta(42) amyloid

81.

82.

83.

84. 85.

86.

87.

88. 89. 90.

91.

92.

93. 94. 95. 96.

observed by cryoEM is relevant at physiological pH. Proc. Natl Acad. Sci. USA, 107, 14128–14133. Popova, L. A., Kodali, R., Wetzel, R. & Lednev, I. K. (2010). Structural variations in the cross-beta core of amyloid beta fibrils revealed by deep UV resonance Raman spectroscopy. J. Am. Chem. Soc. 132, 6324–6328. Zheng, J., Jang, H., Ma, B., Tsai, C. J. & Nussinov, R. (2007). Modeling the Alzheimer A beta(17–42) fibril architecture: tight intermolecular sheet–sheet association and intramolecular hydrated cavities. Biophys. J. 93, 3046–3057. Straub, J. E. & Thirumalai, D. (2011). Toward a molecular theory of early and late events in monomer to amyloid fibril formation. Annu. Rev. Phys. Chem. 62, 437–463. Thirumalai, D., Reddy, G. & Straub, J. E. (2012). Role of water in protein aggregation and amyloid polymorphism. Acc. Chem. Res. 45, 83–92. Sciarretta, K. L., Gordon, D. J., Petkova, A. T., Tycko, R. & Meredith, S. C. (2005). Abeta40-lactam (D23/K28) models a conformation highly favorable for nucleation of amyloid. Biochemistry, 44, 6003–6014. Reddy, G., Straub, J. E. & Thirumalai, D. (2009). Influence of preformed Asp23-Lys28 salt bridge on the conformational fluctuations of monomers and dimers of Abeta peptides with implications for rates of fibril formation. J. Phys. Chem. B, 113, 1162–1172. Tarus, B., Straub, J. E. & Thirumalai, D. (2006). Dynamics of Asp23-Lys28 salt-bridge formation in A beta(10–35) monomers. J. Am. Chem. Soc. 128, 16159–16168. Takeda, T. & Klimov, D. K. (2008). Temperatureinduced dissociation of A beta monomers from amyloid fibril. Biophys. J. 95, 1758–1772. Eichner, T. & Radford, S. E. (2011). A diversity of assembly mechanisms of a generic amyloid fold. Mol. Cell, 43, 8–18. Fawzi, N. L., Ying, J., Ghirlando, R., Torchia, D. A. & Clore, G. M. (2011). Atomic-resolution dynamics on the surface of amyloid-[bgr] protofibrils probed by solution NMR. Nature, 480, 268–272. Grahnen, J. A., Kubelka, J. & Liberles, D. A. (2011). Fast side chain replacement in proteins using a coarse-grained approach for evaluating the effects of mutation during evolution. J. Mol. Evol. 73, 23–33. Buchner, G. S., Murphy, R. D., Buchete, N. V. & Kubelka, J. (2011). Dynamics of protein folding: probing the kinetic network of folding–unfolding transitions with experiment and theory. Biochim. Biophys. Acta, 1814, 1001–1020. Ghosh, A. & Hochstrasser, R. M. (2011). A peptide's perspective of water dynamics. Chem. Phys. 390, 1–13. Humphrey, W., Dalke, A. & Schulten, K. (1996). VMD: Visual Molecular Dynamics. J. Mol. Graphics, 14, 33. Roux, B. & Schulten, K. (2004). Computational studies of membrane channels. Structure, 12, 1343–1351. Jensen, M. O., Tajkhorshid, E. & Schulten, K. (2003). Electrostatic tuning of permeation and selectivity in aquaporin water channels. Biophys. J. 85, 2884–2899.

586 97. Kale, L., Skeel, R., Bhandarkar, M., Brunner, R., Gursoy, A., Krawetz, N. et al. (1999). NAMD2: greater scalability for parallel molecular dynamics. J. Comput. Phys. 151, 283–312. 98. MacKerell, A. D., Bashford, D., Bellott, M., Dunbrack, R. L., Evanseck, J. D., Field, M. J. et al. (1998). Allatom empirical potential for molecular modeling and dynamics studies of proteins. J. Phys. Chem. B, 102, 3586–3616. 99. Martyna, G. J., Tobias, D. J. & Klein, M. L. (1994). Constant-pressure molecular-dynamics algorithms. J. Chem. Phys. 101, 4177–4189. 100. Feller, S. E., Zhang, Y. H., Pastor, R. W. & Brooks, B. R. (1995). Constant-pressure molecular-dynamics

Aβ Protofilament-Lipid Membrane Interactions simulation—the Langevin piston method. J. Chem. Phys. 103, 4613–4621. 101. Jorgensen, W. L., Chandrasekhar, J., Madura, J. D., Impey, R. W. & Klein, M. L. (1983). Comparison of simple potential functions for simulating liquid water. J. Chem. Phys. 79, 926–935. 102. Haass, C., Schlossmacher, M. G., Hung, A. Y., VigoPelfrey, C., Mellon, A., Ostaszewski, B. L. et al. (1992). Amyloid beta-peptide is produced by cultured cells during normal metabolism. Nature, 359, 322–325. 103. Haass, C., Koo, E. H., Mellon, A., Hung, A. Y. & Selkoe, D. J. (1992). Targeting of cell-surface beta-amyloid precursor protein to lysosomes: alternative processing into amyloid-bearing fragments. Nature, 357, 500–503.