Article pubs.acs.org/JPCB
Alzheimer’s Aβ10−40 Peptide Binds and Penetrates DMPC Bilayer: An Isobaric−Isothermal Replica Exchange Molecular Dynamics Study Christopher Lockhart and Dmitri K. Klimov* School of Systems Biology, George Mason University, Manassas, Virginia 20110, United States S Supporting Information *
ABSTRACT: Using all-atom explicit solvent model and isobaric−isothermal replica exchange molecular dynamics, we studied binding of Aβ10−40 monomers to zwitterionic DMPC bilayer. Our simulations suggest three main conclusions. First, binding of Aβ10−40 monomer to the DMPC bilayer causes dramatic structural transition in the peptide resulting in the formation of stable helical structure in the C-terminal. In addition, binding to the lipid bilayer induces the formation of intrapeptide Asp23-Lys28 salt bridge. We argue that the emergence of helix is the consequence of hidden helix propensity harbored in the Aβ10−40 Cterminal. This propensity is revealed by the lipids cross-bridging amino acids in helical conformations and by significant hydrophobic moment of the C-terminal. Second, the central hydrophobic cluster and, particularly, the C-terminal of Aβ10− 40 not only govern binding to the bilayer but also penetrate into bilayer core. In contrast, the polar N-terminal and turn region form interactions mainly with the bilayer surface. Third, our simulations suggest that upon Aβ10−40 binding to the bilayer a highly heterogeneous local environment emerges along Aβ10−40 chain. The N-terminal is exposed to polar well-hydrated medium, whereas the C-terminal is largely shielded from water residing in mostly hydrophobic environment. The implication of our results is that Aβ aggregation mediated by zwitterionic lipid bilayer is likely to be different from that in bulk water.
■
peptide concentrations favor formation of oligomeric species.10 In general, small Aβ oligomers display higher binding affinity compared to large aggregates.12−14 Additionally, anionic bilayers were shown to have higher binding affinity with respect to Aβ than zwitterionic ones.13 According to solution NMR studies, Aβ monomer in water adopts mostly random coil conformations and lacks welldefined native structure.15,16 Similar conclusions follow from all-atom molecular dynamics (MD) simulations, which reveal generally disordered Aβ structural ensemble in water.16−19 It is important to note that both experiments and simulations do not register stable α-helix structure in Aβ C-terminal in water. However, one may expect that the interactions of Aβ with cellular membranes change peptide conformation. For example, helix formation in the C-terminal was observed in several experimental studies probing Aβ structure in membrane-like environments20,21 and in organic solvents.22,23 These observations are consistent with the hidden helix propensity in Aβ Cterminal, which might be revealed by the interactions with the lipid bilayer. Although interactions of Aβ peptides with cellular membranes are likely the key factor in AD pathogenesis, the underlying molecular mechanisms are largely unknown. In recent years, computer simulations have started to address this
INTRODUCTION Aβ peptides, which are the normal products of cellular proteolysis, represent one of the main factors in the development of Alzheimer’s disease (AD).1 Two (β and γ) secretases produce multiple Aβ variants, among which a 40residue peptide, Aβ1−40, is most abundant representing about 90% of all Aβ species in cerebrospinal fluid.2 A large body of in vivo and in vitro experimental evidence suggests that Aβ aggregated forms, particularly oligomers, are the primary cytotoxic species in AD.3,4 However, despite considerable experimental and, more recently, computational efforts, the molecular mechanisms of Aβ cytotoxicity resulting in neuronal death are still uncertain. One of the leading theories suggests that Aβ interactions with cellular membranes play a crucial role in neuronal damage, presumably due to structural perturbation of lipid bilayers and/or formation of pores that disrupt cellular ion homeostasis.5−7 Importantly, Aβ aggregation may occur through different pathways starting in extracellular environment or on/within a cellular membrane. If cytotoxic aggregation is mediated by the membrane, then a monomeric form of membrane-bound Aβ is an initial point in this process and its properties should be examined. Experimental studies supported by theoretical considerations have revealed Aβ association with lipid bilayers.6,8−11 They showed that Aβ interactions with the bilayers are affected by a variety of factors, including Aβ concentration and bilayer composition. Low Aβ concentrations (≲150 nM) result in Aβ binding to the bilayers in monomeric form,11 whereas higher © 2014 American Chemical Society
Received: December 11, 2013 Revised: February 17, 2014 Published: February 18, 2014 2638
dx.doi.org/10.1021/jp412153s | J. Phys. Chem. B 2014, 118, 2638−2648
The Journal of Physical Chemistry B
Article
issue.24 For example, all-atom explicit water MD has been applied to examine the adsorption of Aβ monomers on model lipid bilayers.25−27 MD has been also used to investigate Aβ aggregation in the lipid bilayers. These simulations have suggested that Aβ monomers can form mobile small oligomers in the lipid bilayer, which then assemble into larger, channelforming aggregates.28 However, because computational studies utilize different force fields, membrane models, and sampling methods, they do not yet offer consensus description of Aβ binding to the lipid bilayers. In particular, questions remain concerning the type and extent of structural changes induced in Aβ monomer by binding to the membrane. Furthermore, the mechanism of presumed Aβ penetration into the bilayer has not been firmly established, including the role of different amino acids in binding. In addition, in order to reduce ambiguities attributed to limited sampling, it is imperative to apply simulated tempering methods to study Aβ−bilayer interactions. Motivated by these circumstances, we examine in this study the interactions of Aβ10−40 monomers with zwitterionic DMPC bilayer. As a control we use our previous REMD simulations of Aβ10−40 monomer in water.19 To improve the accuracy of our predictions, we use isobaric−isothermal REMD simulations and explicit solvent all-atom model. Our main results are threefold. First, we demonstrate that Aβ monomer binds with high affinity to the DMPC bilayer and as a result experiences dramatic structural transition by forming stable helix structure in its Cterminal. Second, we found that the central hydrophobic cluster and the C-terminal in Aβ not only govern binding to the bilayer but also penetrate into bilayer core. In contrast, the polar Nterminal and turn region form interactions mainly with the bilayer surface. Third, we show that upon Aβ binding to the bilayer a highly heterogeneous local environment emerges along Aβ chain. The N-terminal occurs in polar well-hydrated medium, whereas the C-terminal is largely shielded from water residing in mostly hydrophobic environment. Implications of our results for the mechanisms of Aβ interactions with cellular membranes are discussed.
Figure 1. (a) Sequence of Aβ10−40 peptide with color-coded regions S1−S4. (b) DMPC lipid is divided into five groups: choline (G1), phosphate (G2), glycerol (G3), and two fatty acid tails (G4 and G5). (c) Simulation system consists of DMPC bilayer (phosphorus atoms at ≈±zP are shown in purple), two Aβ monomers (in cartoon representation), and water (shown by thin lines). (d) Snapshot illustrating insertion of Aβ C-terminal helix into bilayer core. In (c, d) color codes for regions S1−S4 follow those in (a). Terminal Cα atoms are shown as spheres. To improve presentation of panels c and d, a unit cell is flanked on the left and right by its adjacent images.
interval from 8 to 12 Å. Covalent bonds were constrained by the SHAKE algorithm. Two sets of constraints were implemented in the simulation system. The first were the harmonic constraints applied to prevent disintegration of the lipid bilayer at high temperatures (see Replica Exchange Protocol section). The constraints with the force constant k = 6.5 kcal/(mol Å2) approximately fix the center of mass of phosphorus (P) atoms in each leaflet at the distance |zP| = 17.35 Å from the bilayer midplane (z = 0 in Figure 1c). Accordingly, the distance between the centers of mass of P atoms from each leaflet fluctuates around D = 2zP = 34.7 Å. The force constant k and the bilayer thickness D were selected to reproduce the bilayer dimensions and fluctuations observed in the preliminary Aβ-free MD simulations at 330 K. Because the constraints act upon the centers of mass of P atoms, they preserve flexibility of the bilayer. For example, they do not prevent lipids from escaping the bilayer at high temperatures. Furthermore, two constraints independently coupling the centers of mass of P atoms in each leaflet eliminate artificial correlations in their fluctuations. Similar constraints were applied in the simulations of Aβ binding to DPPC and DOPS bilayers.26 Using separate simulations of Aβfree DMPC bilayers at 330 K with and without constraints affecting P atoms, we have verified that these constraints do not affect bilayer structure (as measured by area per lipid or lipid order parameter SCD). The second set is the boundary constraints, which prohibit the peptides from crossing the periodic boundaries along the zdimension. Technically, the boundary constraints were implemented as a pair of repulsive harmonic potentials with the force constant k = 10 kcal/(mol Å2), which act on peptide atom z coordinates, when z is within 4 Å from the periodic boundary. These boundary constraints introduced at the bottom and top of the unit cell prevent interactions of the
■
MODEL AND SIMULATION METHODS All-Atom Explicit Solvent Model. Molecular dynamics (MD) simulations of Aβ monomers and DMPC lipid bilayer were performed using the all-atom explicit solvent CHARMM22 protein force field with CMAP corrections29 and the CHARMM36 lipid force field.30 It has been shown that CMAP corrections improve the consistency between experimental structures and those generated by MD in the disordered regions.29 Following our previous studies,19,31 we used amino-truncated Aβ10−40 peptide as an approximate model of the full length Aβ1−40 (Figure 1a and Supporting Information). DMPC lipid bilayer was selected because its structural and physicochemical properties are well-known.32 The simulation system consisted of two Aβ10−40 monomers interacting with 98 DMPC lipids (Figure 1b,c). Each bilayer leaflet contained 49 lipids arranged in a square shape. Aβ peptides were placed on the opposite sides of the bilayer in the rectangular box of initial dimensions 55.6 Å × 55.6 Å × 81.4 Å containing 4356 water molecules. The protonation states of amino acids corresponded to normal pH. Aβ peptides were capped with neutral acetylated and aminated terminals. Two sodium ions were added to neutralize the system. Simulation system utilized periodic boundary conditions. Electrostatic interactions were computed using Ewald summation, whereas van der Waals interactions were smoothly switched off in the 2639
dx.doi.org/10.1021/jp412153s | J. Phys. Chem. B 2014, 118, 2638−2648
The Journal of Physical Chemistry B
Article
(denoted as ⟨...⟩) were computed using the multiple histogram method37 adapted for NPT simulations.38,39 Its implementation is described in the Supporting Information. Structural quantities related to Aβ represent the averages over two peptides. The results in the article are reported at T = 330 K.
peptides across the periodic cell z boundaries. We checked that the probability for Aβ to interact with the boundary constraints is generally low (0.05). The boundary constraints do not affect water molecules. Replica Exchange Protocol. Because isobaric−isothermal (NPT) ensemble is best suited for simulating lipid bilayers, we have applied NPT replica exchange molecular dynamics (REMD)33 and NAMD program.34 Compared to the more commonly used canonical REMD,35 NPT REMD requires certain modifications. Consider a potential energy of the system E(q), where q are atomic coordinates. The replica i of the system is described by the state xim,n, where the indexes m and n correspond to the temperature Tm and pressure Pn. NPT MD simulations are performed in parallel for all replicas, each characterized by the unique combination of T and P. Monte Carlo simulations in the replica space attempt to exchange the replicas i and j with the states xim,n and xjm+1,n+1 as xim,n → xjm,n and xjm+1,n+1 → xim+1,n+1. The probability ω of accepting the exchange follows from the detailed balance condition ω = min[1, exp(− (Δ1 + Δ2 ))]
■
RESULTS Aβ Conformational Ensemble. Using NPT REMD, we have analyzed the equilibrium conformational ensemble of Aβ10−40 monomer interacting with zwitterionic DMPC bilayer at 330 K (Figure 1c). At this temperature, the probability of peptide binding to the bilayer Pb (as defined in the Supporting Information) is ≈1.0. The distribution of Aβ secondary structure presented in Figure 2 shows the fractions of helix ⟨H(i)⟩, turn ⟨T(i)⟩, and random coil ⟨RC(i)⟩ for each amino acid i. Averaged over the entire peptide, the fractions ⟨H⟩, ⟨T⟩, and ⟨RC⟩ are 0.39 ± 0.03, 0.30 ± 0.03, and 0.31 ± 0.01, respectively. The contributions from other secondary structure types are less than 0.01. Thus, the largest contribution to the conformational ensemble of the bound peptide is provided by helix. To analyze the distribution of secondary structure along the sequence, we assume that a secondary structure is stable at i if the respective fraction (⟨H(i)⟩, ⟨T(i)⟩, or ⟨RC(i)⟩) exceeds 0.5 (Figure 2). Then, the most stable secondary structure in Aβ is helix in the C-terminal (residues 31−37, region S4), whereas a short, less stable helix emerges in S3 (residues 23−26). Random coil is frequently observed at the positions 16−20 (mostly S2 region) and 27−29 spanning across S3 and S4. Turn is sampled at the positions 12−14 in S1. Other sequence regions have either no stable structure or it involves less than three amino acids. Consistent conclusions follow, if we analyze the secondary structure in Aβ regions S1− S4 (Table 1). This table reveals the following distribution of secondary structure in Aβ monomer: a combination of turn and random coil in S1, random coil in S2, a combination of helix and turn in S3, and stable helix in S4. Aβ peptide tertiary structure is explored by the intrapeptide contact map ⟨C(i, j)⟩, which gives the equilibrium probabilities of forming side chain contacts between residues i and j (j > i + 1). Figure 3a reveals numerous contacts in bound Aβ peptide with the total number of intrapeptide contacts ⟨C⟩ = 28.4 ± 0.7. Stable contacts (⟨C(i, j)⟩ > 0.5) can be divided into turnforming (i to i + 2 interactions), helix-forming (i to i + 3 or i to i + 4 interactions), or long-range interactions (j − i ≥ 5; because π-helix is rare, we assigned i to i + 5 interactions to long-range category). The turn-forming interactions include Ser26-Lys28 (0.76), Gly25-Asn27 (0.60), and Gly37-Val39 (0.57). The interactions favoring helix appear in or near the segments S4 and S3 and include Gly33-Val36 (0.88), Gly33Gly37 (0.83), Ala21-Val24 (0.74), Ile32-Met35 (0.63), Gly37Val40 (0.62), Ala30-Gly33 (0.62), Ile31-Leu34 (0.58), Phe20Val24 (0.57), Gly29-Ile32 (0.52), and Ala21-Gly25 (0.51). Asp23-Lys28 salt bridge (0.79) is the only stable long-range contact, which is also the third strongest among all intrapeptide interactions. Formation of Asp23-Lys28 contact is illustrated by computing the distribution of distances between the side chains of Asp23 and Lys28, rKD (Figure 3b). This distribution reflects strong electrostatic attractive interactions between Asp23 and Lys28, resulting in small value of ⟨rKD⟩ = 5.6 ± 0.4 Å. Aside from Asp23-Lys28 contact, there are several other long-range contacts with significant stability (⟨C(i, j)⟩ > 0.35; Figure 3a): Phe19-Ile31 (0.48), Val24-Ile31 (0.42), and Phe19-Leu34 (0.35). These contacts represent hydrophobic interactions
(1)
where Δ1 = (βm − βm+1)(E(q ) − E(q )), Δ2 = (βmPn − βm+1Pn+1)(Vn+1 − Vn), β = 1/kBT, and Vn and Vn+1 are the volumes of the replicas i and j. If all replicas are simulated at the same constant pressure P = Pn, then the indexes m and n in eq 1 can be replaced with one index m. In REMD simulations the pressure was held constant by applying the Langevin piston method,36 and a semi-isotropic pressure coupling was adopted to adjust the coupled x and y box dimensions (in the bilayer plane) independently of the zdimension. Temperature was controlled through underdamped Langevin simulations of ”virtual” solvent with the damping coefficient γ = 5 ps−1. The integration step was set to 1 fs. We have distributed R = 40 replicas exponentially in the temperature range from 320 to 430 K. Also, we have chosen to simulate all replicas at the same constant pressure P = 1 atm. Exchanges were attempted every 2 ps between all neighboring replicas along the temperature scale generating an average acceptance rate of 24%. We produced five independent REMD trajectories resulting in the cumulative simulation time of 4 μs or 100 ns per each replica. Because initial portions of REMD trajectories are not equilibrated, the cumulative equilibrium simulation time was reduced to 3.3 μs. It is of note that the simulation system includes two Aβ peptides, which according to our analysis do not interact. Consequently, effective sampling time per peptide is 6.6 μs. REMD trajectories were started with Aβ peptides in different states, including unbound and bound peptides sampling random coil, turn, and helix conformations. The DMPC bilayer was pre-equilibrated. Analysis of REMD convergence is presented in the Supporting Information. Computation of Structural Probes. Molecular interactions were computed between individual amino acids or Aβ sequence regions and the bilayer. To this end, Aβ10−40 peptide was divided into hydrophilic N-terminal (S1, residues 10−16), central hydrophobic cluster (S2, residues 17−21), hydrophilic turn (S3, residues 22−28), and hydrophobic Cterminal (S4, residues 29−40) (Figure 1a). A DMPC lipid molecule was divided into five structural groups: choline (G1), phosphate (G2), glycerol (G3), and two fatty acid tails (G4 and G5) (Figure 1b). Computation of interactions in Aβ-bilayer system and structural quantities are described in the Supporting Information. Thermodynamic averages of structural quantities j
i
2640
dx.doi.org/10.1021/jp412153s | J. Phys. Chem. B 2014, 118, 2638−2648
The Journal of Physical Chemistry B
Article
between amino acids Phe19 from the central hydrophobic cluster (S2) or Val24 from the turn (S3) with Ile31 or Leu34 from the hydrophobic C-terminal (S4). Aβ−Bilayer Interactions. To characterize interactions of individual amino acids with the bilayer, we first plot in Figure 4a the probabilities of binding Pb(i) for amino acids i. (An amino acid is bound, if it forms at least one contact with the lipids as defined in the Supporting Information.) We classify an amino acid i as binding if Pb(i) > 0.5. It follows from Figure 4a that there are four binding amino acids in the central hydrophobic cluster S2, Leu17 (Pb(i) = 0.53), Val18 (0.69), Phe20 (0.79), and Ala21 (0.71). The turn S3 contains three binding amino acids, Val24 (0.60), Gly25 (0.68), and Asn27 (0.71), but the largest number of binding residues (11, in all) are located in the C-terminal S4, which include Gly29 (0.63), Ala30 (0.57), Ile31 (0.51), Ile32 (0.65), Gly33 (0.55), Met35 (0.59), Val36 (0.73), Gly37 (0.64), Gly38 (0.70), Val39 (0.65), and Val40 (0.71). No binding amino acids are found in the Nterminal S1. Table 2 lists the average values of binding probabilities for Aβ regions. Because their binding probabilities are in excess of 0.5, the regions S2 and S4 are classified as binding. Therefore, Figure 4a and Table 2 suggest that two hydrophobic regions of Aβ peptide, central hydrophobic cluster S2 and the C-terminal S4, play crucial role in Aβ−bilayer interactions. The turn region S3 contributes to binding to a lesser extent, whereas the hydrophilic N-terminal S1 participates weakly in Aβ−bilayer interactions. Further insight into Aβ−bilayer interactions is provided by the probabilities Pc(i) for amino acids i to interact with the bilayer core, i.e., to form at least one contact with the fatty acid groups G4 and G5 (see Supporting Information for contact definition). The values of Pc averaged over the regions S1−S4 (Table 2) indicate that overall S2 and S4 have the largest probabilities of interacting with fatty acids, i.e., penetrating into bilayer core. Figure 4a demonstrates that the amino acids with the largest Pc(i)(>0.5) are Phe20 (0.58), Ile32 (0.58), Met35 (0.52), Val36 (0.61), Val39 (0.54), and Val40 (0.56), of which the majority (five out of six) are from the C-terminal S4 and one from S2. Two other hydrophobic amino acids, Val24 (from S3) and Ile31 (from S4), also have large Pc(i)(>0.40). Similar results follow if we consider the number of contacts ⟨Cl(i,k)⟩ between amino acid i and the lipid group k (= G1, ..., G5) (Figure 4b). Indeed, nearly all interactions with fatty acids (G4 and G5) are attributed to the regions S2 or S4. According to Table 2, these two regions form by far the largest number of contacts with fatty acids compared to S1 or S3. In particular, the largest number of interactions with fatty acids are formed by Phe20 from S2 and Ile32, Val36, Val39, and Val40 from S4, all of which are hydrophobic residues. Importantly, Table 2 and Figure 4b also reveal interactions of S2 and S4 with the surface lipid groups G1−G3. In contrast, the regions S1 and S3 interact weakly with the core lipid groups forming almost exclusively the contacts with G1−G3. Among the amino acids favoring interactions with the surface lipid groups are polar Lys16 and Asn27, which form the contacts Asn27-G1 (⟨Cl(i,k)⟩ = 0.68), Lys16-G2 (0.65), and Asn27-G2 (0.58). Formation of hydrogen bonds between Aβ peptide and bilayer is analyzed in the Supporting Information. Taken together, Figure 4 and Table 2 indicate that the central hydrophobic cluster S2 and the C-terminal S4 not only govern binding to the bilayer but are also responsible (especially, S4) for penetration into bilayer core. In contrast, polar N-terminal S1 and turn S3 form interactions mainly with
Figure 2. Distribution of secondary structure in Aβ peptide with respect to amino acids i: (a) helix ⟨H(i)⟩; (b) turn ⟨T(i)⟩; (c) random coil ⟨RC(i)⟩. Data for the Aβ−bilayer system is in black, for Aβ coincubated with ibuprofen or in water are in red and blue, respectively. Color codes for regions S1−S4 follow Figure 1a. The plot shows that binding to the DMPC bilayer causes profound shift in Aβ structure manifested mainly by the formation of helix in the Cterminal. Inset to (a): the fraction of amino acids in helical conformation ⟨H(z)⟩ at the distance z from the bilayer center. The position of the center of mass of phosphorus atoms in bilayer leaflet is marked by the dashed line. The plot shows that as a peptide draws closer to the bilayer center helix content increases. A small dip near z ∼ 20 Å is due to that most amino acids found at this z come from the N-terminal S1, which has low helix propensity. Note that ⟨H(z)⟩ at large z does not represent the water bulk value because even at those z the peptide is still partially bound to the bilayer. 2641
dx.doi.org/10.1021/jp412153s | J. Phys. Chem. B 2014, 118, 2638−2648
The Journal of Physical Chemistry B
Article
Table 1. Secondary Structure in Aβ Peptide environment DMPC bilayer
waterb
ibuprofen solutionb
a
secondary structure ⟨H⟩ ⟨T⟩ ⟨RC⟩ ⟨H⟩ ⟨T⟩ ⟨RC⟩ ⟨H⟩ ⟨T⟩ ⟨RC⟩
S1a 0.14 0.46 0.40 0.04 0.41 0.55 0.17 0.39 0.44
± ± ± ± ± ± ± ± ±
S2a
0.08 0.06 0.04 0.01 0.02 0.01 0.03 0.03 0.02
0.10 0.39 0.50 0.02 0.59 0.38 0.09 0.49 0.42
± ± ± ± ± ± ± ± ±
S3a
0.03 0.03 0.01 0.00 0.01 0.01 0.00 0.02 0.02
0.39 0.41 0.20 0.16 0.65 0.19 0.34 0.44 0.22
± ± ± ± ± ± ± ± ±
0.05 0.06 0.02 0.01 0.02 0.01 0.02 0.01 0.01
S4a 0.65 0.10 0.26 0.17 0.42 0.40 0.51 0.22 0.27
± ± ± ± ± ± ± ± ±
0.05 0.05 0.01 0.02 0.02 0.01 0.02 0.02 0.00
In Tables 1−3 S1−S4 are Aβ regions defined in Figure 1a. bData from ref 19.
Figure 4. (a) Probabilities of binding to the DMPC bilayer, Pb(i), (in gray) and probabilities of binding to lipid fatty acid groups G4 and G5, Pc(i), (in black) for each amino acid i. The value of Pb = 0.5 used to define binding amino acids is marked. The plot shows that the central hydrophobic cluster (the region S2) and the C-terminal (S4) are primarily responsible for binding to the DMPC bilayer. (b) Contact map ⟨Cl(i,k)⟩ visualizes the number of contacts between amino acid i and lipid group k. This map suggests that the regions S2 and S4 interact with the core (k = G4 and G5) and surface (G1−G3) lipid groups, whereas the interactions of S1 and S3 are restricted to the surface lipid groups. Color codes for regions S1−S4 follow Figure 1a.
Figure 3. (a) Intrapeptide contact map ⟨C(i, j)⟩ visualizes the probability of interactions between the amino acids i and j (|j − i| >1). Color codes for regions S1−S4 follow Figure 1a. (b) Probability distribution P(rKD) for the distance between the side chains of Asp23 and Lys28 rKD. Data in black, red, and blue are for Aβ bound to the bilayer, Aβ coincubated with ibuprofen, and Aβ in water, respectively. In contrast to water environment, binding to the DMPC or coincubation with ibuprofen induce the formation of the Asp23Lys28 salt bridge. The inset shows a bilayer bound Aβ with formed salt bridge between Asp23 (in red) and Lys28 (in blue).
the probability distributions P(z;i) of occurrence of amino acids i along the z-axis. This plot strikingly suggests that most Cterminal amino acids are inserted into the bilayer because for them the maximum P(z;i) occurs at z < zP, where zP = 17.35 Å is the average position of the center of mass of phosphorus atoms. The regions S2 and, to a lesser extent, S3 reveal similar behavior for some amino acids, such as Phe20 from S2 or Val24 from S3. In contrast, the amino acids from the N-terminal S1 are almost never inserted into the bilayer as their z > zP. This
the bilayer surface. These points are further elaborated in the next section. Aβ Insertion into Bilayer. The results presented above suggest that some Aβ amino acids penetrate the bilayer core. In order to directly investigate this possibility, we plot in Figure 5a 2642
dx.doi.org/10.1021/jp412153s | J. Phys. Chem. B 2014, 118, 2638−2648
The Journal of Physical Chemistry B
Article
Table 2. Aβ−Bilayer Interactions S1 interaction probabilitiesa
contactsc
Pb Pc Ps Pi G1−G3d G4, G5e
0.33 0.02 0.45 0.09 0.22 0.01
± ± ± ± ± ±
S2 0.04 0.00 0.06 0.03 0.03 0.00
0.62 0.38 0.28 0.52 0.26 0.30
± ± ± ± ± ±
S3 0.13 0.13 0.07 0.18 0.05 0.10
0.46 0.11 0.46 0.30 0.27 0.07
± ± ± ± ± ±
S4 0.09 0.04 0.05 0.10 0.06 0.02
0.61 ± 0.13 0.38 ± 0.11 0.12 ± 0.04 0.66b ± 0.19 0.21 ± 0.04 0.29 ± 0.08
a Average value per amino acid. bFor S4 Pi >Pb, because folded and inserted structure in S4 precludes some amino acids from interaction with the bilayer. cMeasured by the average number of contacts ⟨Cl(i,k)⟩ between amino acid from Aβ region and lipid group. The data are normalized by the number of lipid groups in each category. dLipid surface groups. eLipid fatty acid groups.
Nx(i) for amino acids in S1−S4 regions. It is also convenient to define the difference ΔN(i) = Nap(i) − Npl(i), which characterizes local environment around amino acid i (Figure 6). Table 3 and Figure 6 demonstrate that S1 is in a largely polar (ΔN(i) < 0) and well-hydrated environment, and there are few lipid atoms in the vicinity of S1. In contrast, the S2 environment is largely hydrophobic (ΔN(i) > 0) and compared to S1 dehydrated (Nw for S2 is about half of S1). Because the local lipid atom number Nl is tripled compared to S1, S2 strongly interacts with the bilayer. The environment of S3 is similar to S1 characterized by relatively small Nl and elevated number of water atoms Nw. In Table 3 the number of polar atoms Npl near S3 is almost equal to that for S1 and well above Nap. Consequently, according to Figure 6, all S3 amino acids except one are in polar environment (ΔN(i) < 0). Finally, according to Table 3, the S4 environment is the most dehydrated and the local lipid atom number Nl is similar to that observed for S2. Figure 6 shows that ΔN(i) > 0 for almost all S4 amino acids; i.e., the S4 environment is highly hydrophobic, in which Nap reaches maximum among all Aβ regions (exceeding, e.g., that for S1 by a factor of 2). Table 3 also shows that the local number of peptide atoms progressively increases from the N- to C-terminal, reflecting ascending degree of Aβ folding. Thus, the analysis of Figure 6 and Table 3 suggests that upon binding the N-terminal (S1) and, to a lesser extent, the turn region (S3) reside in polar well-hydrated environment. In contrast, the central hydrophobic cluster S2 and, especially, the C-terminal S4 mainly exist in apolar dehydrated medium. These results are consistent with the analysis of binding and insertion of Aβ into the DMPC bilayer. Computations of accessible surface areas for amino acids presented in the Supporting Information (Figure S6) complement our conclusions concerning local environment.
distinctive behavior is illustrated for several amino acids in Figure S5. In Figure 5b we further analyze peptide distribution along the z-axis by computing the probabilities for each amino acid i to be near the bilayer surface Ps(i) or inserted into the bilayer Pi(i). An amino acid is assumed bound to the bilayer surface, if its z coordinate is within zP < z < 24.8 Å. (The upper boundary corresponds to the maximum extent of binding interactions for amino acids along the z-axis.) An amino acid is considered inserted in the bilayer, if z < zP. With these definitions the amino acids classified as inserted (Pi(i) > 0.5) are Val18, Phe20, Ala21, Val24, and all C-terminal residues 29− 40, confirming our interpretation of Figure 5a above. Of all amino acids, only His14 and Gln15 from the N-terminal S1 and Glu22 and Asn27 from S3 have the values of Ps(i) greater than 0.5. If the probabilities of insertion Pi and surface binding Ps are computed for Aβ regions (Table 2), then S1 and S3 have the highest probabilities of surface binding, whereas S2 and, particularly, S4 have the highest probability of insertion into the bilayer (both >0.5). This analysis directly confirms the conclusion from the previous section that the central hydrophobic cluster S2 and, particularly, the C-terminal S4 penetrate the bilayer core as illustrated in Figure 1d. Finally, peptide insertion can be characterized by plotting the probability distributions n(z) for peptide, lipid, and water atoms as a function of z (Figure 5c). This plot reveals a bimodal distribution of peptide atoms along the bilayer leaflet with two maxima at z ≃ 13 and 25 Å located, respectively, below and above the layer of phosphorus atoms at zP. Therefore, Figure 5c reinforces our conclusion that Aβ amino acids exist in two states, in which they interact with the bilayer: (1) bound to the bilayer surface and (2) inserted into the bilayer. Both states coexist in Aβ molecule with the regions S1 and S3 being largely surface bound and the regions S2 and S4 being mostly inserted in the bilayer core. Figure 5c also suggests that inserted amino acids occur in generally dehydrated environment, where the lipid atom density reaches its peak. Aβ Binding Environment. Because of different binding propensities of Aβ regions, one may expect that these regions experience different local environments. To characterize those around Aβ, we computed several radial distribution functions (rdf) gx(r;i) for each amino acid i. Specifically, the functions gw(r;i), gp(r;i), and gl(r;i) measure the local number densities of water, peptide, and lipid heavy atoms, respectively, at the distance r from amino acid i. In addition, we defined gpl(r;i) and gap(r;i), which measure the local number densities of polar and apolar atoms from any molecule. Using these functions, we computed the local atom numbers Nx(i) = ∫ 6.5 0.0gx(r;i)v(r) dr, where v(r) is volume element. By definition, Nx(i) is the number of atoms of the type x in the sphere of the radius of 6.5 Å around amino acid i. Table 3 provides the average values of
■
DISCUSSION Does Binding to the DMPC Bilayer Change Aβ Conformational Ensemble? Using all-atom explicit solvent model and isobaric−isothermal REMD simulations, we have studied the conformational ensemble of Aβ monomer bound to zwitterionic DMPC bilayer. We found that the largest contribution to Aβ secondary structure is provided by helix, which represents about 40% of conformational ensemble. Analysis of the secondary structure along Aβ sequence in Figure 2a and Table 1 indicates that stable helix occurs in the C-terminal (residues 31−37, region S4) and within the turn region S3 (23−26). In addition, a short but stable turn structure appears at the positions 12−14 in the N-terminal (region S1). To evaluate the changes in Aβ structure induced by binding to the DMPC bilayer, we use as a reference our previous 2643
dx.doi.org/10.1021/jp412153s | J. Phys. Chem. B 2014, 118, 2638−2648
The Journal of Physical Chemistry B
Article
Figure 6. Atom number difference ΔN(i) = Nap(i) − Npl(i) probes local environment around amino acid i. Positive and negative ΔN(i) correspond to mostly hydrophobic and polar environments, respectively. The data in black, red, and blue are for Aβ monomers bound to the bilayer, coincubated with ibuprofen, and in water. Color codes for regions S1−S4 follow Figure 1a.
Table 3. Local Environment around Bound Aβ Monomer S1 Np Nl Nw Npl Nap
10.9 2.9 13.0 18.4 8.5
± ± ± ± ±
S2 0.2 0.4 0.2 0.1 0.2
14.2 8.1 6.3 12.5 16.1
± ± ± ± ±
S3 0.6 2.1 1.4 1.3 1.4
15.9 4.6 9.9 17.8 12.6
± ± ± ± ±
S4 0.1 1.1 0.7 0.3 0.7
16.9 7.8 5.4 12.8 17.3
± ± ± ± ±
0.2 1.9 1.4 1.2 1.5
2, which identifies three stable turn regions, encompassing residues 19−25, 27−29, and 35−37. Compared to the bilayer bound state the fraction of helix ⟨H⟩ in water is reduced three times (from 0.39 to 0.13). The turn fraction ⟨T⟩ is elevated from 0.30 to 0.48, whereas the random coil fraction experiences minor changes (0.31 vs 0.38). According to Figure 2 and Table 1, the most dramatic structural difference is observed in the Cterminal, where the helix fraction reaches ⟨H(S4)⟩ = 0.65 in Aβ bound to the bilayer but drops 4-fold to 0.17 in water. Similar but less dramatic changes are observed in the turn region S3, where ⟨H(S3)⟩ = 0.39 vs 0.16 in water. Computation of the root-mean-square deviation (RMSD) between the helix distributions in water and in the bilayer-bound Aβ over the entire sequence results in the large value of 0.37. Computations of the tertiary interactions in Aβ monomer bound to the bilayer show that stable turn and helix forming interactions appear in the C-terminal S4 and the turn S3 regions (Figure 3a). None of these interactions (apart from Gly37-Val39 and Ala21-Val24) occur in Aβ conformations in water.19 In the bilayer bound Aβ Asp23-Lys28 salt bridge is the only stable long-range interaction formed with the probability of 0.79 (inset to Figure 3b). The average distance separating Asp23 and Lys28 is 5.6 Å. In contrast, in water Asp23-Lys28 salt bridge is disrupted as its probability is reduced to 0.17 and the average distance ⟨rDK⟩ is increased more than 2-fold to 11.9 Å (Figure 3b). Consequently, the average RMSD between the intrapeptide contact maps for the peptides bound to the bilayer and in water reaches the value of 0.09. It is also instructive to compare the conformational ensembles of Aβ monomers bound to the DMPC bilayer and coincubated with ibuprofen ligands.19 The latter samples helical
Figure 5. (a) Probability distributions P(z;i) of occurrence of amino acid i at the position z along the normal to the DMPC bilayer. The striking feature is bimodal localization of amino acids, suggesting that they fall into two categories, bound to the bilayer surface (or remaining unbound in water) and inserted into the bilayer. (b) Probabilities of insertion into the bilayer Pi(i) (in black) and binding to its surface Ps(i) (in gray) for amino acids i. The plot reveals the preference for the C-terminal (S4) and the central hydrophobic cluster S2 to be inserted into the bilayer. In (a) and (b) color codes for regions S1−S4 follow Figure 1a. (c) Probability distributions n(z) along the normal to the DMPC bilayer for peptide, lipid, and water atoms shown in black, red, and blue, respectively. Consistent with (a) this plot reveals bimodal distribution of peptide atoms. Dashed lines in (a) and (c) indicate the positions of center of mass of phosphorus atoms from bilayer leaflet. The value z = 0 corresponds to the bilayer midplane.
REMD simulations of Aβ monomer in lipid-free water.19 The secondary structure in Aβ monomer in water is shown in Figure 2644
dx.doi.org/10.1021/jp412153s | J. Phys. Chem. B 2014, 118, 2638−2648
The Journal of Physical Chemistry B
Article
originates from the hidden helix propensity harbored in Aβ Cterminal. In our recent work we have investigated the conformational ensembles of several amino-truncated Aβ peptides in water. 40 We showed that the equilibrium distributions of structures adopted by Aβ23−40 and Aβ10− 40 are similar but sharply distinct from the conformational ensemble of Aβ29−40. This peptide fragment, which exactly coincides with the region S4, forms a stable helical structure not present in longer fragments. Therefore, the helix propensity in Aβ10−40 C-terminal, which is hidden in water, can be revealed by ligand binding or by binding to the zwitterionic bilayer. The mechanism revealing this propensity is in part based on crossbridging contacts formed by either ligands or lipids biasing the peptide toward helical state. Finally, we compare the local environments established around Aβ peptides in different systems. Figure 6 shows that the environment for bilayer bound Aβ is more hydrophobic than for Aβ coincubated with ibuprofen. For bilayer bound Aβ, 16 residues including all in S2, Val24 of S3, and residues 31−40 of S4 have positive values of ΔN(i), indicating that there are more apolar than polar atoms in their vicinity. For comparison, Aβ coincubated with ibuprofen contains six residues with positive values of ΔN(i) (Phe19 of S2, Val24 of S3, and Ile31, Gly33, Leu34, and Met35 of S4). As seen in Figure 6, the difference in environment hydrophobicity is even more pronounced when bilayer bound Aβ is compared against the peptide in water, which contains no residues with positive ΔN(i). Thus, all residues in Aβ in water are in polar environment, whereas Aβ bound to the bilayer or coincubated with ibuprofen feature many residues (mostly from S4) in an apolar environment. Our main result that binding to the zwitterionic DMPC bilayer induces helix structure in Aβ monomer is consistent with several prior studies. Experiments have probed Aβ structure in membrane mimicking environments, such as SDS micelles.20,21 They observed the formation of α-helix in the sequence regions 15−24 and 29−35. Interestingly, the helix in the region 15−24 is destabilized by the increase in pH above 6.0, but the C-terminal helix remains largely unaffected.20 Such distribution of secondary structure is in a good agreement with the one in Figure 2. Similar observations concerning the distribution of helical structure in Aβ peptides have been made for organic solvents, such as TFE or fluorinated alcohol.22,23 We are aware of two computational studies, which utilized REMD to study Aβ−bilayer interactions. REMD simulations and simplified representation of lipid bilayer based on continuum dielectric model have been used to analyze the structure of Aβ1−40 monomer in the bilayer.41 That study has revealed that interactions with the membrane induce helix structure in the sequence regions 13−25 and 30−35. These results are consistent (especially for 30−35 region) with our findings. The second study used canonical (NVT) REMD and united atom model to explore binding of Aβ monomer to zwitterionic DPPC bilayer, which is structurally similar to DMPC.26 The authors did not observe formation of any specific secondary structure in the peptide bound to the bilayer nor stabilization of Asp23-Lys28 salt bridge. It remains to be seen if the difference between our results is due to different force fields or to different REMD ensembles (NVT vs our NPT). Constant temperature MD simulations reaching 100 ns at 323 K have been performed to probe the structural dynamics of Aβ1−40 monomer preinserted into DPPC bilayer in helix-rich
(⟨H⟩ = 0.33), turn (⟨T⟩ = 0.35), and random coil (⟨RC⟩ = 0.32) conformations and forms two regions with stable secondary structure (Figure 2). The first (residues 20−25) is dominated by turn and helix structure, whereas a stable helix emerges in the second region (29−35). These findings as well as Figure 2 and Table 1 indicate that the secondary structure (particularly, helix) distributions in Aβ monomers bound to the DMPC bilayer and coincubated with ibuprofen are strikingly similar. As a result, the RMSD between the helix distributions in Aβ bound to the bilayer and coincubated with ibuprofen is reduced to 0.16, which is more than twice smaller than the corresponding RMSD comparing Aβ peptides bound to the bilayer and in water. In simulations of Aβ with ibuprofen, helix stabilization was attributed to ligands cross-bridging residues i to i + 3 or i to i + 4 adopting helical conformation.19 For instance, in the helixforming C-terminal 68% of ligand contacts were involved in cross-bridging. Similar computations for bilayer bound Aβ show that 60% of lipid contacts with S4 residues are involved in cross-bridging. Corresponding fraction for S3, which also forms significant helix structure in Aβ−bilayer system, is 49%, but for S1 and S2, which do not form helix, it drops to 25% and 37%. Taken together, these results suggest that similar to ibuprofen solution lipid cross-bridging is expected to stabilize helix structure in bilayer bound Aβ. Our previous study has found that Aβ peptide coincubated with ibuprofen forms extensive network of stable intrapeptide interactions.19 There are four stable turn-forming contacts, of which three (Gly25-Asn27, Ser26-Lys28, and Gly37-Val39) also occur in the peptide bound to the bilayer. Aβ peptide coincubated with ibuprofen contains seven stable helix-forming contacts, of which five are also found in the peptide bound to the bilayer (Phe20-Val24, Ala21-Val24, Gly29-Ile32, Ile31Leu34, and Gly33-Val36). Finally, similar to bilayer-bound Aβ there is only one stable long-range contact in the peptide coincubated with ibuprofen, Asp23-Lys28, which is formed with the probability of 0.53 and ⟨rKD⟩ = 7.4 Å (Figure 3b). Consequently, the RMSD comparing the contact maps for the peptides bound to the bilayer and coincubated with ibuprofen is 0.05, which is reduced almost in half from the RMSD comparing Aβ bound to the bilayer and in water. Several conclusions follow from the comparative analysis above. First, binding to the bilayer causes dramatic restructuring in Aβ monomer reflected (i) in emergence of extensive helix structure in the C-terminal S4 and, to a lesser extent, in the turn region S3 and (ii) in formation of stable Asp23-Lys28 salt bridge (Figure 3b). Using REMD sampling, we can show directly that binding to the bilayer induces helix structure in Aβ. To this end, we plot in the inset to Figure 2a the fraction of Aβ helix structure ⟨H(z)⟩ as a function of the distance z between the peptide center of mass and the bilayer midplane. The figure shows that as Aβ penetrates deeper into the bilayer below the layer of phosphorus atoms at zP a steady increase in ⟨H(z)⟩ is observed. Thus, the conformations of Aβ bound to the zwitterionic bilayer are strikingly different from those sampled in water. This observation suggests that the aggregation of Aβ peptides mediated by the bilayer should follow different pathway than in bulk water. Second, somewhat unexpectedly we found similarities in the conformational ensembles of Aβ peptides bound to the bilayer and coincubated with ibuprofen. Both conformational ensembles feature stable helix structure in the C-terminal and stable Asp23-Lys28 salt bridge. We believe that this similarity 2645
dx.doi.org/10.1021/jp412153s | J. Phys. Chem. B 2014, 118, 2638−2648
The Journal of Physical Chemistry B
Article
S3) with the bilayer are lower than −14 kcal/mol (for comparison, other electrostatic or van der Waals bilayer−amino acid interactions are at least twice weaker). These conclusions are supported by Figure 6, which shows all S2 residues and residues 31−40 of S4 in apolar environments, whereas all S1 and most of S3 residues are in polar medium. Additionally, the distribution of rASA in Figure S6 also shows the regions S2 and S4 as shielded from water. Of the two regions, S2 and S4, which penetrate the bilayer, only the C-terminal S4 exhibits stable helix structure, whereas S2 forms mostly random coil (Figure 2). The presence of helix is visible in Figure 5a, where S4 amino acids show sawtooth-like oscillations in their most probable positions along z. Similar oscillations are not observed in the unstructured S2. These oscillations imply that upon insertion the position of Cterminal helix is effectively locked. As a result residues Gly29, Ala30, Ile31, Gly33, Leu34, Gly37, Gly38, and Val40 have larger z values than Ile32, Met35, Val36, and Val39. Spacial separation of glycines from other hydrophobic amino acids creates a hydrophobic moment, which according to MPEx tool46 is 5.1 μH for the C-terminal residues 29−37. For comparison, the hydrophobic moments for other Aβ regions are 1.3 (S1), 1.9 (S2), and 2.6 μH (S3). Previous studies indicated that large hydrophobic moments promote helix formation in membrane bound antimicrobial and amyloidogenic peptides.47 We believe that, after the cross-bridging contacts discussed above, the existence of significant hydrophobic moment is the second reason for revealing the hidden helix propensity in the C-terminal. It is important to compare the interaction mechanism emerging from our simulations with those reported in the computational literature. REMD simulations of Aβ1−40 interacting with implicit lipid bilayer have studied the location of the peptide in the bilayer.41 That study has shown that Aβ1− 40 tends to reside at the membrane−water interface. Furthermore, the C-terminal was found to be partially inserted in the bilayer being tilted at the angle γ = 135° with respect to the bilayer normal (Figure 1c). For comparison, Figure S7 presents the probability distribution P(γ) obtained from our simulations. The distribution is bimodal with the largest peak occurring at γ ≈ 135° and the smaller one at 60°. The location of the largest peak is consistent with the value obtained in ref 41. Constant temperature MD simulations using explicit solvent and lipid models revealed that the C-terminal of Aβ1−40 monomer remains embedded in the zwitterionic DPPC bilayer for 100 ns anchored by strong electrostatic interactions between Lys28 and the lipid headgroups.25,43 Similar conclusion has been reached using 200 ns explicit water simulations, which started with Aβ structures, in which Cterminal is embedded perpendicularly in the bilayer surface.48 The latter simulations showed that the peptide gradually moves to a conformation parallel to the surface.25,48,49 Thus, the published reports are in agreement with our REMD simulations that Aβ C-terminal favors shallow insertion. However, the previous studies differ from our results concerning the role of Lys28.25,48 While in previous simulations a positively charged Lys28 forms electrostatic interactions with the lipid groups, our data suggest that Lys28 instead prefers to form stable intrapeptide salt bridge with Asp23. Indeed, the total energies of interaction between Lys28 and peptide, lipid, or water atoms are −115.6, −7.7, and −60.0 kcal/mol, respectively. This demonstrates that intrapeptide interactions involving Lys28 are
conformations.25 It was found that Aβ C-terminal partially retains helix structure on the simulation time scales, particularly if the peptide is inserted deeper in the bilayer. Longer (500 ns) simulations have reported even stronger helix stability in the Cterminal for Aβ1−42 monomer inserted in the DPPC bilayer.42 Interestingly, the same study showed that preformed βstructure in Aβ1−42 monomer is largely unstable in the DPPC bilayer. Another recent report has shown that the fraction of helix structure increases as a function of insertion into the bilayer. 43 This result supports our REMD computations, which reveal the same trend in helix population as a function of z (inset to Figure 2a). Finally, it is important to mention REMD simulations applied to investigate binding of the model peptide WALP-16 to DPPC bilayer.44 This study suggested a novel mechanism of peptide binding to the bilayer, in which helix acquisition follows (but not precedes) binding and, most importantly, insertion into the bilayer core. According to the inset to Figure 2a, this scenario applies to our REMD simulations of Aβ interactions with the DMPC bilayer. Mechanism of Aβ Interaction with Zwitterionic Bilayer. Our REMD simulations allow us to investigate equilibrium mechanism of interaction of Aβ monomer with zwitterionic DMPC bilayer. By computing the probability of binding for each amino acid Pb(i), we classify as binding four amino acids from the central hydrophobic cluster S2, three from the turn region S3, and all amino acids except Leu34 from the C-terminal S4. To get further insight into Aβ−bilayer interactions, we computed the interactions of amino acids with individual lipid groups, including the probability of interacting with the bilayer core, Pc(i). On the basis of these computations (presented in Figure 4 and Table 2), we concluded that the central hydrophobic cluster S2 and the C-terminal S4 not only govern binding to the bilayer but are the regions (especially, S4), which penetrate into bilayer core. Polar N-terminal S1 and turn S3 demonstrate different behavior by forming interactions mainly with the bilayer surface. To verify these conclusions, we have analyzed the distribution of amino acids along the axis z normal to the bilayer surface, P(z;i). Figure 5a,b and Figure S5 suggest that amino acids interact with the bilayer by binding to its surface or by penetrating into its core. This two-state interaction mode is reflected in the bimodal probability distribution for peptide atoms in Figure 5c. The amino acids classified as inserted (from Figure 5b) are Val18, Phe20, and Ala21 from S2, Val24 from S3, and all Cterminal residues 29−40. Analysis with respect to Aβ regions in Table 2 shows that S1 and S3 have the highest probabilities of surface binding, whereas S2 and, particularly, S4 have the highest probability of insertion into the bilayer (Figure 1d). Importantly, both of these regions are exclusively composed of hydrophobic amino acids, whereas S1 and S3 are highly hydrophilic (six out of seven in S1 and five out of seven amino acids in S3 are polar with S1 and S3 having positive and negative net charges). Because the C-terminal S4 and central hydrophobic cluster S2 have the largest probabilities of interacting with the bilayer (Figure 4a and Table 2), hydrophobic effect appears to be the main factor in Aβ interactions with the zwitterionic DMPC bilayer leading to partial burial of the peptide. However, consistent with the recent study45 electrostatic interactions also play important role in binding the hydrophilic regions S1 and S3 to the bilayer surface. For example, the average energies of electrostatic interactions of Glu11, Lys16 (both from S1), or Glu22 (from 2646
dx.doi.org/10.1021/jp412153s | J. Phys. Chem. B 2014, 118, 2638−2648
The Journal of Physical Chemistry B
■
far more favorable than the interactions with lipids or water. This is consistent with the recent conclusion that Lys28−lipid interactions and formation of salt-bridge Asp23-Lys28 are in direct competition.26 As a result, according to Figure 4a Lys28 plays minor role in binding to the bilayer. These discrepancies are likely to arise due to different simulation strategies. While previous simulations utilized constant temperature sampling, our study used REMD to enhance conformational sampling. Several experimental studies have probed the interactions of Aβ peptides with lipid bilayers. A combination of CD spectroscopy, electron microscopy, and AFM was used to study binding of Aβ1−40 peptides to DMPC bilayer.8 It was concluded that the peptide partially inserts through the Cterminal 29−40 region into the bilayer. A similar picture emerged from NMR and CD experiments,9 suggesting that the degree of insertion of Aβ C-terminal into the bilayer depends on the charged state of the lipid. Zwitterionic lipids (such as DMPC) favor surface binding or shallow penetration into the bilayer, whereas anionic lipids facilitate deeper insertion of Aβ into bilayer core. These experimental observations are consistent with our study, but they also point out that the mechanism of Aβ−bilayer interaction described by us is likely to be restricted to zwitterionic lipids. In addition, our work does not address the question of Aβ aggregation mediated by the lipid bilayer. Experiments have found that Aβ peptides bind as monomers at low peptide concentrations (≲ 150 nM).11,50 Thus, the mechanism of Aβ−bilayer interaction described by us may differ from those observed at elevated peptide concentrations or for anionic lipids.
■
Article
ASSOCIATED CONTENT
S Supporting Information *
Further details concerning the model and methods used and additional data. This material is available free of charge via the Internet at http://pubs.acs.org.
■
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected] (D.K.K.). Notes
The authors declare no competing financial interest.
■
REFERENCES
(1) Hardy, J.; Selkoe, D. J. The Amyloid Hypothesis of Alzheimer’s Disease: Progress and Problems on the Road to Therapeutics. Science 2002, 297, 353−356. (2) Haass, C.; Schlossmacher, M. G.; Hung, A. Y.; Vigo-Pelfrey, C.; Mellon, A.; Ostaszewski, B. L.; Lieberburg, I.; Koo, E. H.; Schenk, D.; Teplow, D. B.; et al. Amyloid β-Peptide Is Produced by Cultured Cells During Normal Metabolism. Nature 1992, 359, 322−325. (3) Haass, C.; Selkoe, D. J. Soluble Protein Oligomers in Neurodegeneration: Lessons from the Alzheimers Amyloid β-Peptide. Nat. Rev. Mol. Cell. Biol. 2007, 8, 101−112. (4) Shankar, G. M.; Li, S.; Mehta, T. H.; Garcia-Munoz, A.; Shepardson, N. E.; Smith, I.; Brett, F. M.; Farrell, M. A.; Rowan, M. J.; Lemere, C. A.; et al. Amyloid-β Protein Dimers Isolated Directly from Alzheimers Brains Impair Synaptic Plasticity and Memory. Nat. Med. 2008, 14, 837−842. (5) Williams, T. L.; Serpell, L. C. Membrane and Surface Interactions of Alzheimer’s Aβ Peptide - Insights into the Mechanism of Cytotoxicity. FEBS J. 2011, 278, 3905−3917. (6) Arispe, N.; Diaz, J. C.; Simakova, O. Aβ Ion Channels. Prospects for Treating Alzheimer’s Disease with Aβ Channel Blockers. Biochim. Biophys. Acta 2007, 1768, 1952−1965. (7) Nakazawa, Y.; Suzuki, Y.; Williamson, M.; Saito, H.; Asakura, T. the Interaction of Amyloid Aβ(1−40) with Lipid Bilayers Ganglioside As Studied by 31P Solid State NMR. Chem. Phys. Lipids 2009, 158, 54− 60. (8) Yip, C.; McLaurin, J. Amyloid-β Peptide Assembly: A Critical Step in Fibrillogenesis and Membrane Disruption. Biophys. J. 2001, 80, 1359−1371. (9) Bokvist, M.; Lindstrom, F.; Watts, A.; Grobner, G. Two Types of Alzheimer’s β-Amyloid (1−40) Peptide Membrane Interactions: Aggregation Preventing Transmembrane Anchoring Versus Accelerated Surface Fibril Formation. J. Mol. Biol. 2004, 335, 1039−1049. (10) Qulst, A.; Doudevski, I.; Lin, H.; Azimova, R.; Ng, D.; Franglone, B.; Kagan, B.; Ghiso, J.; Lal, R. Amyloid Ion Channels: A Common Structural Link for Protein-Misfolding Disease. Proc. Natl. Acad. Sci. U. S. A. 2005, 102, 10427−10432. (11) Nag, S.; Chen, J.; Irudayaraj, J.; Maiti, S. Measurement of the Attachment and Assembly of Small Amyloid-β Oligomers on Live Cell Membranes at Physiological Concentrations Using Single-Molecule Tools. Biophys. J. 2010, 99, 1969−1975. (12) Cizas, P.; Budvytyte, R.; Morkuniene, R.; Moldovan, R.; Broccio, M.; Losche, M.; Niaura, G.; Valincius, G.; Borutatite, V. SizeDependent Neurotoxicity of β-Amyloid Oligomers. Arch. Biochem. Biophys. 2010, 496, 84−92. (13) Kremer, J. J.; Murphy, R. M. Kinetics of Adsorption of βAmyloid Peptide Aβ(1−40) to Lipid Bilayers. J. Biochem. Biophys. Methods 2003, 57, 159−169. (14) Cizas, P.; Budvytyte, R.; Morkuniene, R.; Modovan, R.; Broccio, M.; Losche, M.; Niaura, G.; Valincius, G.; Borutaite, V. SizeDependent Neurotoxicity of β-Amyloid Oligomers. Arch. Biochem. Biophys. 2010, 496, 84−92. (15) Hou, L.; Shao, H.; Zhang, Y.; Li, H.; Menon, N. K.; Neuhaus, E. B.; Brewer, J. M.; Byeon, I.-J. L.; Ray, D. G.; Vitek, M. P.; et al. Solution NMR Studies of the Aβ(1−40) and Aβ(1−42) Peptides
CONCLUSIONS
Using all-atom explicit solvent model and novel NPT REMD simulations, we have studied the conformational ensemble of Aβ monomer bound to zwitterionic DMPC bilayer. There are three main conclusions in our study. First, binding of Aβ monomer to the DMPC bilayer causes dramatic structural transition in the peptide, resulting in formation of stable helix in the C-terminal. Another consequence of binding to the bilayer is the formation of intrapeptide Asp23-Lys28 salt bridge. We found that these structural changes are remarkably similar to those induced by ibuprofen binding probed in our previous simulations. We argued that the emergence of helix in the Cterminal is the consequence of hidden helix propensity harbored in the Aβ C-terminal. This propensity is revealed by the lipids cross-bridging amino acids in helical state and by significant hydrophobic moment of the C-terminal. Second, the central hydrophobic cluster and, particularly, the C-terminal of Aβ not only govern binding to the bilayer but also penetrate into the bilayer core. In contrast, the polar N-terminal and turn regions form interactions mainly with the bilayer surface. Third, we show that upon Aβ peptide binding to the bilayer a highly heterogeneous local environment emerges along Aβ chain. The N-terminal retains polar well-hydrated environment, whereas the C-terminal is largely shielded from water residing in mostly hydrophobic medium. We argued that our findings are in reasonable agreement with previous experimental and computational studies. The implication of our results is that Aβ aggregation mediated by zwitterionic lipid bilayer is likely to be different from that in bulk water. 2647
dx.doi.org/10.1021/jp412153s | J. Phys. Chem. B 2014, 118, 2638−2648
The Journal of Physical Chemistry B
Article
Establish That the Met35 Oxidation State Affects the Mechanism of Amyloid Formation. J. Am. Chem. Soc. 2004, 126, 1992−2005. (16) Sgourakis, N.; Merced-Serrano, M.; Boutsidis, C.; Drineas, P.; Du, Z.; Wang, C.; Garcia, A. Atomic-Level Characterization of the Ensemble of the Aβ(1−42) Monomer in Water Using Unbiased Molecular Dynamics Simulations and Spectral Algorithms. J. Mol. Biol. 2011, 405, 570−583. (17) Yang, M.; Teplow, D. B. Amyloid β-Protein Monomer Folding: Free Energy Surfaces Reveal Alloform Specific Differences. J. Mol. Biol. 2008, 384, 450−464. (18) Anand, P.; Hansmann, U. H. E. Internal and Environmental Effects on Folding and Dimerization of the Alzheimer’s Beta-Amyloid Peptide. Mol. Simul. 2011, 37, 440−448. (19) Lockhart, C.; Kim, S.; Klimov, D. K. Explicit Solvent Molecular Dynamics Simulations of Aβ Peptide Interacting with Ibuprofen Ligands. J. Phys. Chem. B 2012, 116, 12922−12932. (20) Coles, M.; Bicknell, W.; Watson, A. A.; Fairlie, D. P.; Craik, D. J. Solution Structure of Amyloid Beta-Peptide(1−40) in a Water-Micelle Environment. Is the Membrane Spanning Domain Where We Think It Is? Biochemistry 1998, 37, 11064−11077. (21) Jarvet, J.; Danielsson, J.; Damberg, P.; Oleszczuk, M.; Graslund, A. Positioning of the Alzheimer Aβ(1−40) Peptide in SDS Micelles Using NMR and Paramagnetic Probes. J. Biomol. NMR 2007, 39, 63− 72. (22) Sticht, H.; Bayer, P.; Willbold, D.; Dames, S.; Hilbich, C.; Beyreuther, K.; Frank, R. W.; Rosch, P. Structure of Amyloid A4-(1− 40)-Peptide of Alzheimer’s Disease. Eur. J. Biochem. 1995, 233, 293− 298. (23) Crescenzi, O.; Tomaselli, S.; Guerrini, R.; Salvadori, S.; D’Ursi, A. M.; Temussi, P. A.; Picone, D. Solution Structure of the Alzheimer’s Amyloid β-Peptide (1−42) in an Apolar Microenvironment. Similarity with a Virus Fusion Domain. Eur. J. Biochem. 2002, 269, 5642−5648. (24) Ma, B.; Nussinov, R. Simulations As Analytical Tools to Understand Protein Aggregation and Predict Amyloid Conformation. Curr. Opin. Struct. Biol. 2006, 10, 445−452. (25) Lemkul, J.; Bevan, D. A Comparative Molecular Dynamics Analysis of the Amyloid β-Peptide in a Lipid Bilayer. Arch. Biochem. Biophys. 2008, 470, 54−63. (26) Davis, C.; Berkowitz, M. Structure of the Amyloid-β (1−42) Monomer Absorbed to Model Phospholipid Bilayers: A Molecular Dynamics Study. J. Phys. Chem. B 2009, 113, 14480−14486. (27) Wang, Q.; Zhao, J.; Yu, X.; Zhao, C.; Li, L.; Zheng, J. Alzheimer Aβ1−42 Monomer Adsorbed on the Self-Assembled Monolayers. Langmuir 2010, 26, 12722−12732. (28) Jang, H.; Arce, F. T.; Ramachandran, S.; Capone, R.; Azimova, R.; Kagan, B. L.; Nussinov, R.; Lal, R. Truncated β-Amyloid Peptide Channels Provide an Alternative Mechanism for Alzheimer Disease and Down Syndrome. Proc. Natl. Acad. Sci. U. S. A. 2010, 107, 6538− 6543. (29) Buck, M.; Bouguet-Bonnet, S.; Pastor, R. W.; MacKerell, A. D. Importance of the CMAP Correction to the CHARMM22 Protein Force Field: Dynamics of Hen Lysozyme. Biophys. J. 2006, 90, L36− L38. (30) Klauda, J.; Venable, R.; Freites, J.; O’Connor, J.; Tobias, D.; Mondragon-Ramirez, C.; Vorobyov, I.; A.D. MacKerell, J.; Pastor, R. Update of the CHARMM All-Atom Additive Force Field for Lipids: Validation on Six Lipid Types. J. Phys. Chem. B 2010, 114, 7830−7843. (31) Lockhart, C.; Klimov, D. Molecular Interactions of Alzheimer’s Biomarker FDDNP with Aβ Peptide. Biophys. J. 2012, 103, 2341− 2351. (32) Nagle, J. F.; Tristram-Nagle, S. Structure of Lipid Bilayers. Biochim. Biophys. Acta 2000, 1469, 159−195. (33) Okabe, T.; Kawata, M.; Okamoto, Y.; Mikami, M. Replica Exchange Monte-Carlo Method for the Isobaric-Isothermal Ensemble. Chem. Phys. Lett. 2001, 335, 435−439. (34) Kale, L.; Skeel, R.; Bhandarkar, M.; Brunner, R.; Gursoy, A.; Krawetz, N.; Phillips, J.; Shinozaki, A.; Varadarajan, K.; Schulten, K. NAMD2: Greater Scalability for Parallel Molecular Dynamics. J. Comput. Phys. 1999, 151, 283−312.
(35) Sugita, Y.; Okamoto, Y. Replica-Exchange Molecular Dynamics Method for Protein Folding. Chem. Phys. Lett. 1999, 114, 141−151. (36) Feller, S.; Zhang, Y.; Pastor, R.; Brooks, B. Constant Pressure Molecular Dynamics Simulation: the Langevin Piston Method. J. Chem. Phys. 1995, 103, 4613−4621. (37) Ferrenberg, A.; Swendsen, R. Optimized Monte Carlo Data Analysis. Phys. Rev. Lett. 1989, 63, 1195−1198. (38) Chang, J.; Sandler, S. Determination of Liquid-Solid Transition Using Histogram Reweighting Method and Expanded Ensemble Simulations. J. Chem. Phys. 2003, 118, 8390−8395. (39) Conrad, P.; de Pablo, J. Comparison of Histogram Reweighting Techniques for a Flexible Water Model. Fluid Phase Equilib. 1998, 150−151, 51−61. (40) Lockhart, C.; Klimov, D. Revealing Hidden Helix Propensity in Aβ Peptides by Molecular Dynamics Simulations. J. Phys. Chem. B 2013, 117, 12080−12088. (41) Miyashita, N.; Straub, J.; Thirumalai, D. Structures of β-Amyloid Peptide 1−40, 1−42, and 1−55-the 672−726 Fragment of APP-in a Membrane Environment with Implications for Interactions with γSecretase. J. Am. Chem. Soc. 2009, 131, 17843−17852. (42) Poojari, C.; Kukol, A.; Strodel, B. How the Amyloid-β Peptide and Membranes Affect Each Other: An Extensive Simulation Study. Biochim. Biophys. Acta 2012, 1828, 327−339. (43) Lemkul, J.; Bevan, D. Lipid Composition Influences the Release of Alzheimer’s Amyloid β-Peptide from Membranes. Protein Sci. 2011, 20, 1530−1545. (44) Nymeyer, H.; Woolf, T.; Garcia, A. Folding Is Not Required for Bilayer Insertion: Replica Exchange Simulations of an α-Helical Peptide with an Explicit Lipid Bilayer. Proteins: Struct., Funct., Bioinf. 2005, 59, 783−790. (45) Davis, C. H.; Berkovitz, M. L. Interaction Between Amyloidβ(1−42) Peptide and Phospholipid Bilayer: A Molecular Dynamics Study. Biophys. J. 2009, 96, 785−797. (46) Snider, C.; Jayasinghe, S.; Hristova, K.; White, S. MPEx: A Tool for Exploring Membrane Proteins. Protein Sci. 2009, 18, 2624−2628. (47) Last, N. B.; Miranker, A. D. Common Mechanism Unites Memrane Poration by Amyloid and Antimicrobial Peptides. Proc. Natl. Acad. Sci. U. S. A. 2013, 110, 6382−6387. (48) Qiu, L.; Buie, C.; Reay, A.; Vaughn, M.; Cheng, K. Molecular Dynamics Simulations Reveal the Protective Role of Cholesterol in βAmyloid Protein-Induced Membrane Disruption in Neuronal Membrane Mimics. J. Phys. Chem. B 2011, 115, 9795−9812. (49) Xu, Y.; Shen, J.; Luo, X.; Zhu, W.; Chen, K.; Ma, J.; Jiang, H. Conformational Transition of Amyloid β-Peptide. Proc. Natl. Acad. Sci. U. S. A. 2005, 102, 5403−5407. (50) Ding, H.; Schauerte, J.; Steel, D.; Gafni, A. β-Amyloid (1−40) Peptide Interactions with Supported Phospholipid Membranes: A Single-Molecular Study. Biophys. J. 2012, 103, 1500−1509.
2648
dx.doi.org/10.1021/jp412153s | J. Phys. Chem. B 2014, 118, 2638−2648