Article pubs.acs.org/JPCB

Interactions of a Water-Soluble Fullerene Derivative with Amyloid‑β Protofibrils: Dynamics, Binding Mechanism, and the Resulting SaltBridge Disruption Xiaoying Zhou,† Wenhui Xi,† Yin Luo, Siqin Cao, and Guanghong Wei* State Key Laboratory of Surface Physics, Key Laboratory for Computational Physical Sciences (Ministry of Education), Department of Physics, Fudan University, 220 Handan Road, Shanghai 200433, People’s Republic of China S Supporting Information *

ABSTRACT: Alzheimer’s disease (AD) is associated with the pathological self-assembly of amyloid-β (Aβ) peptides into β-sheet-rich oligomers and insoluble amyloid fibrils. Experimental studies reported that 1,2-(dimethoxymethano)fullerene (DMF), a watersoluble fullerene derivative, inhibits strongly Aβ peptide aggregation at the early stage. However, the interaction and binding mechanisms are not well understood. In this study, we have investigated the detailed interaction of a DMF molecule with a fibrillar hexamer of full-length Aβ42 and the resulting structural alterations by performing multiple all-atom explicit solvent molecular dynamics (MD) simulations. Starting from different initial states with a minimum distance of 2 nm between the DMF and the Aβ protofibril, our MD simulations show that the DMF binds to the Aβ protofibril via both slow and fast binding processes. Three dominant binding sites are identified: the central hydrophobic core (CHC) site (17LVFFA21), the turn site (27NKGAI31), and the C-terminal β-sheet site consisting of the smallest side-chain residue glycine and hydrophobic residues (31IIGLMVGGVVI41). Binding energy analyses reveal the importance of π-stacking interactions, especially in the CHC site, hydrophobic interactions, and curvature matching. Strikingly, we find that the binding of DMF to the turn region can disrupt the D23−K28 salt-bridge that is important for the Aβ fibrillation. These results provide molecular insight into the binding mechanism of fullerene to Aβ protofibrils and offer new routes for the therapeutic drug design using fullerene derivatives against AD.

1. INTRODUCTION The aggregation of amyloidogenic proteins is associated with many serious human diseases, including Alzheimer’s disease (AD), Parkinson’s disease (PD), and type II diabetes.1 Among these, AD is the most common neurodegenerative disorder, which is characterized by fibrillar deposits of amyloid-β (Aβ) peptides and neurofibrillary tangles of Tau proteins.2,3 Aβ is a 39- to 43-residue-long peptide, derived from the amyloid precursor protein (APP) by the proteolytic cleavage of β- and γ-secretases.4 The predominant components of the fibrillar deposits in the brains of AD patients are 40- and 42-residuelong Aβ peptides (Aβ40 and Aβ42).5 X-ray diffraction data show that the fibrillar deposits consist of fibrils that display a cross-β structure with the β-strands perpendicular and the interstrand hydrogen bonds (H-bonds) parallel to the fibril axis.6 The fibril structures of Aβ peptides have been proposed based on the solid-state nuclear magnetic resonance (NMR) data and MD simulations.7,8,29 The fibrillation process can be described by a nucleation-elongation process characterized by a lag phase associated with the formation of a variety of toxic intermediate oligomeric states such as small oligomers and protofibrils, after which the fibril growth proceeds rapidly.9,10 Through twisting and wrapping with each other, the protofibrils join together to form polymorphic fibrils, which are also cytotoxic to neurons.11,12 © 2014 American Chemical Society

The aggregation process can be affected by many factors such as metal ions,13 membranes,14 or nanoparticles (NPs).15,16 The carbon-based NPs, including fullerenes (C60)17−19 and carbon nanotubes (CNTs),20 have received considerable attention in the context of AD. However, the poor solubility of carbon NPs in water has been a major hindrance in terms of their potential biomedical applications. The water-soluble derivatives of C60 fullerenes have been synthesized and are suggested to be useful in the treatment of neurodegenerative diseases.17,21−24 For example, in vitro experiments showed that 1,2(dimethoxymethano)fullerene (DMF), a fullerene derivative (a more soluble form of fullerene), inhibits strongly the Aβ peptide aggregation at the early stage.18 Other water-soluble fullerene C60 derivatives were also found to be able to inhibit amyloid fibrillation and reduce the cytotoxity of Aβ peptides.17,22 These experimental studies have greatly enhanced our understanding of the influence of fullerene derivatives on the aggregation and the toxicity of Aβ peptide; however, the interaction dynamics, binding mechanism, and resulting structural alterations of Aβ during the Aβ−fullerene interaction process are not well understood. Received: April 8, 2014 Revised: May 23, 2014 Published: May 23, 2014 6733

dx.doi.org/10.1021/jp503458w | J. Phys. Chem. B 2014, 118, 6733−6741

The Journal of Physical Chemistry B

Article

Figure 1. (a) The structure of DMF. (b) The initial conformation of the Aβ42 protofibril with a DMF molecule placed at three different positions. In the three initial states, the minimum distance between DMF and Aβ42 is 2 nm. For DMF molecules, carbon atoms are in gray, oxgen atoms are in red, and CH3 (united-atom) is in green. For Aβ42, different colors represent different residues types: blue, positively charged; red, negatively charged; green, polar; gray, hydrophobic. Four different surfaces of Aβ protofibril are defined as follows: elongation surface (with main chain atoms exposed to solvent), β1 surface (with the side chains of even residues 2, 4, ..., 22 exposed to solvent), β2 surface (with the side chains of odd residues 29, 31, ..., 35 exposed to solvent), and N-terminal surface (with the side chains of odd residues 1, 3, ..., 11 exposed to solvent).

through the curvature-matched binding as well as hydrophobic interactions. In addition, the hydrophobic concave on the surface of the Aβ hexamer at the CHC region also facilitates DMF molecule binding. This binding would block the backbone amide sites for further monomer or oligomer addition with β-sheet structure and inhibits or slows down the elongation process. Binding to the turn region can disrupt the salt-bridge formation that is important for fibril formation. These results provide molecular insight into the binding mechanism of DMF molecules to Aβ protofibrils and the resulting structural changes of protofibrils.

Complementary to experimental studies, all-atom molecular dynamics (MD) simulations can provide atomic-level information on the interactions between fullerenes and amyloid peptides. Using docking method and MD simulations, Andujar et al. investigated the effect of a pristine fullerene on the structure of an idealized pentameric construct of Aβ units (a model fibril). Three different binding sites were identified, and two of them were located in the polar, disordered region, which was thought to be the irrelevant ones.25 The third one, which is located in the core part of the pentamer, was chosen as the most relevant binding site. Their MD simulations, starting from the Aβ−C60 complex with fullerene binding to the third binding site, showed that this binding caused significant structural alterations of the Aβ model fibril.25 However, the binding dynamics, binding mechanism, and binding-induced structural changes by DMF, a water-soluble fullerene derivative, remain elusive. The purpose of the present work is to investigate the interactions between a DMF molecule and a Aβ42 protofibril at the atomic level using all-atom explicit-solvent MD simulations. Following the work by Wu and Shea,26 we choose a hexamer as a protofibril model in our MD simulations. We have carried out multiple MD runs starting from an Aβ42 protofibril with a DMF molecule placed 2.0 nm away from the protofibril. These MD simulations demonstrate that DMF molecules bind to multiple binding sites of Aβ protofibrils within a 90 ns time scale. Three primary binding sites are identified: the central hydrophobic core (CHC) (17LVFFA21) site, the turn site (27NKGAI31), and the C-terminal β-sheet site (31IIGLMVGGVVI41). The CHC binding site found here is consistent with the one reported previously in an experimental study27 and an computational study using the docking method.25 The turn site and the C-terminal β-sheet site are new binding sites and were not reported previously. The interaction analysis of Aβ− DMF reveals two predominant binding mechanisms. The first mechanism involves a π-stacking interaction of aromatic residues with DMF molecules, and the second is mainly

2. MATERIALS AND METHODS DMF Molecule. The molecular structure of DMF used in this study is shown in Figure 1a. The force field parameters were taken from a previous study28 and are given in Table S1 (Supporting Information). An united-atom model is employed for DMF molecules; therefore, the methyl group is represented by a sphere (see Figure 1a). Fibrillar Aβ42 Hexamer. Due to the polymorphisms of Aβ fibril, several fibril structure models have been proposed based on the solid-state NMR data and MD simulations.29,30 Here, we use the Ma−Nussinov−Tycko model of the Aβ42 fibril structure29 as a starting state to extract the fibrillar Aβ42 hexamer (see Figure S1(a), Supporting Information), in which the N-terminal residues 1−16 are in a β-sheet structure. It is noted that the N-terminal residues 1−16 are missing in most of the Aβ fibril models.7 The coordinate of the full-length Aβ42 protofibril is kindly provided by Ma and Nussinov. The Aβ42 monomer in the protofibril consists of an N-terminal βstrand (β1) (residues 2−22), a C-terminal β-strand (β2) (residues 30−42), and a turn (residues 23−29) (see Figure 1b). The Aβ 42 hexamer is energy-minimized and properly equilibrated (see Figure 1b and Figure S1(a), Supporting Information). In order to describe the binding of DMF to Aβ42 protofibril more conveniently, we define four different surfaces for the solvent-exposed surface of Aβ42 protofibril (see Figure 6734

dx.doi.org/10.1021/jp503458w | J. Phys. Chem. B 2014, 118, 6733−6741

The Journal of Physical Chemistry B

Article

Figure 2. Time evolution of the minimum distance dmin between DMF and Aβ42 starting from the three different initial states I, II, and III shown in Figure 1b. For each initial state, three representative trajectories (1, 2, 3) are given (a) for I, (b) for II, and (c) for III.

1b): the N-terminal surface (odd residues 1, 3, ..., 11), β1 surface (even residues 2, 4, ..., 22), β2 surface (odd residues 29, 31, ..., 35), and elongation surfaces (the backbone of the first and the sixth chains in the protofibril). DMF−Aβ Protofibril System. The DMF−Aβ protofibril system consists of a fibrillar Aβ42 hexamer and a DMF molecule placed 2.0 nm (minimum distance) away from the hexamer. In order to remove the bias of the initial position of the DMF molecule on the binding site, the DMF molecule is initially placed at three different locations (I, II, III) (see Figure 1b). Each DMF−Aβ protofibril system is placed in a rectangular box of water molecules with a minimum distance to the box wall of 2 nm. As a control, the Aβ hexamer without DMF is also simulated in two independent 50 ns MD runs. Simulation Details. The MD simulations have been carried out in the isothermal−isobaric (NPT) ensemble using the GROMACS 4.5.3 software package31 with the GROMOS96 43a1 force field.32 Water molecules are represented by the simple point charge (SPC) model.33 Bond lengths of both the peptides and DMF molecules are constrained with the LINCS algorithm,34 and the water geometry is constrained with the SETTLE algorithm,35 which allow an integration time step of 2 fs. Long-range electrostatic interactions are calculated using the particle mesh Ewald (PME) method with a real space cutoff of 0.9 nm.36 The van der Waals (vdW) interactions are calculated by using a cutoff of 1.4 nm. The solute and solvent are separately coupled to the external temperature and pressure baths using, respectively, velocity-rescale37 and Parrinello− Rahman methods.38 All of the MD simulations have been performed at a temperature of 310 K and a pressure of 1 bar. The coupling times of the temperature and pressure are 0.1 and 1.0 ps, respectively. Ten independent MD runs have been performed starting from each initial state shown in Figure 1b. As the time for DMF bound to Aβ varies in different MD runs, the lengths of MD runs are different. The MD runs are 10−90 ns long, and the last 5 ns of data is used for analysis. The simulation time for the control system without DMF is 50 ns, and the last 5 ns trajectory is used for data analysis. Analysis. Data analyses have been performed using the GROMACS facilities and our in-home-developed codes. A binding site is considered if the distance between the DMF molecule and the residues in one region of Aβ protofibril maintains within 0.35 nm for at least 5 ns. In the calculation of the minimum distance between Aβ and DMF, only heavy atoms of Aβ are considered. We analyze the aromatic stacking interactions between Aβ and DMF by calculating the probability density function (PDF) of the distance between

the centroids of the aromatic ring of Phe residues and the closest carbon ring of the DMF molecule and the PDF of the angle between these two rings. The D23−K28 distance is the distance between the centroids of the COO− and NH3+ groups in their side chains. Snapshots are shown using the VMD program.39 The DMF−protofibril complex structures generated in the last 5 ns MD runs are clustered using the Daura method implemented in GROMACS. The binding energy (in units of kcal/mol) of DMF at the different sites of the Aβ hexamer is estimated by using the MM/GBSA methodology as implemented in the AMBER package40 using the same force field as above. The free energy of a system in the MM/GB(PB)SA is computed as the sum of the molecular mechanics energy in the gas phase, the solvation free energy, and the conformational entropy. This approach is computationally less expensive than free-energy perturbation and thermodynamic integration methods as it considers only the unbound and bound states. This method is widely used to implicitly consider the solvation free energy for a complex system.41 The generalized Born (GB) model developed by Onufriev and co-workers42 is employed to estimate the solvent effect. The conformational entropies of proteins have not been calculated by MM/GB(PB)SA as done previously by others.43−45 The energy components of each residue are also calculated using the idecomp method.46 The curvature of the molecular surface for the Aβ protofibril is estimated by the Surface Racer software developed by Sergeev and co-workers.47 The probe radius is set to be 0.50 nm, comparable with the radius of the fullerene (about 0.34 nm) and the vdW radius of the carbon atom on the surface of the Aβ hexamer.

3. RESULTS AND DISCUSSION Before investigating the binding behavior of a DMF molecule to the fibrillar Aβ hexamer, we examine the structural property of Aβ hexamer in the absence of DMF molecules by conducting two independent 50 ns MD simulations at 310 K. Visual inspection of the equilibrated structures in the two MD runs reveals that the hexamer can largely retain the strand−turn− strand conformation seen in Aβ fibrils.29 Structural analysis of the Aβ hexamer in one of the two MD runs is given in Figure S1 (Supporting Information). Due to the flexibility of residues 37GG38 and the strong hydrophobicity of 39VVIA42, in the equilibrated protofibril, the C-terminal β-sheet spanning residues 39VVIA42 bend outward, as shown in Figure S1(a) (Supporting Information). Similar conformations were also observed in the fibrillar Aβ17−42/Aβ42 pentamers reported previously.25,48 The bending of the C-terminal β-sheet indicates 6735

dx.doi.org/10.1021/jp503458w | J. Phys. Chem. B 2014, 118, 6733−6741

The Journal of Physical Chemistry B

Article

Table 1. Binding Sites and the Corresponding Binding Energy ΔGbindinga binding sites

number of md runs

N-terminal nonaromatic residues 4FRH6 8SGY10 13HH14 CHC (17LVFFA21) Turn (27NKGAI31) C-terminal β-sheet Others

5 2 4 2 6 5 2 4

preferred surface β1 surface β1 surface elongation surface β2 surface β2 surface

ΔGbinding −19.2 −21.5 −27.9 −24.4 −32.2 −34.5 −34.9 −21.81

± ± ± ± ± ± ± ±

3.9 3.3 4.6 5.4 4.8 5.8 4.3 3.3

ΔEvdW

ΔEelect

ΔGgB

ΔGSurf

−25.1 −25.5 −30.4 −26.4 −34.5 −37.1 −36.0 −24.8

−2.5 −1.0 −0.5 −2.6 −1.5 −2.9 −0.6 −1.6

10.8 7.4 6.2 7.0 7.2 8.6 5.2 7.1

−2.7 −2.5 −3.3 −2.5 −3.5 −3.2 −3.6 −2.6

The number of trajectories leading to each binding site is also shown. ΔGbinding = ΔEvdW + ΔEelect + ΔGgB + ΔGSurf. Data are shown as the mean with the standard deviation. The different components of the binding energy are also given. ΔEvdw, van der Waals potential energy; ΔEelect, electrostatic potential energy; ΔGgB, electrostatic contributions to the solvation free energy; ΔGSurf, nonpolar contributions to the solvation free energy. Mean values are calculated from the different trajectories for each binding site resulting from independent simulations. The unit of energy is kcal/mol. a

hexamer without DMF is also based on the last 5 ns trajectory. We also examine the influence of the initial position of DMF on the binding sites. The binding sites found in the 30 different MD runs are given in Table S2 (Supporting Information). We observe that starting from the same initial position, the DMF molecule can find different binding sites; starting from different positions, the DMF molecule can find the same binding site. These results indicate that the binding sites are not biased by the initial positions of the DMF molecule. After characterizing the binding processes of DMF to the Aβ protofibril, we examine whether the dimethoxymethane group of DMF has a preference to interact with the Aβ protofibril. To this aim, we present, in Figure S2 (Supporting Information) the time evolution of the minimum distance (dmin) between the dimethoxymethane group and the protofibril in three representative trajectories after the DMF molecule binds to the protofibril (only 5 ns of data is shown). The time point for the DMF molecule starting to bind to the protofibril is different in the three different trajectories. It can be seen from Figure S2 (Supporting Information) that, depending on the initial velocity distribution of the MD run, after the DMF molecule binds to the protofibril, the dmin changes continuously or occasionally between 0.35 and 0.95 nm within 5 ns, indicating that the DMF molecule can rotate at the binding site, especially in runs I-1 and II-1 shown in Figure S2(a,b) (Supporting Information). The rotation behavior of the DMF is similar to that of pristine fullerenes bound to an Aβ pentamer reported in a recent computational study25 but less frequent than the pristine fullerene. In particular, in 2 out of 30 MD trajectories, the DMF molecule is observed to rotate only one time within 5 ns (Figure S2(c), Supporting Information). These results imply that the introduction of the dimethoxymethane group on the fullerene surface can retard the rotation of the fullerene, thus enhancing the binding stability of the fullerene. Three Primary Binding Sites Identified. Identifying the binding sites of a DMF molecule on the surface of the Aβ protofibril is the first step to understand its binding mechanism. We show in Figure S3 (Supporting Information) all of the positions of bound DMF on the surface of the Aβ protofibril in all 30 MD runs. As discussed above, the binding sites are not biased by the initial positions of the DMF molecule as it is placed 2 nm away from the protofibril (i.e., there are no molecular interactions between them initially). It can be seen from Figure S3 (Supporting Information), the DMF molecules visited a wide range of the protofibril surface. However, the binding sites are not distributed homogeneously on the

that the lateral association might be favorable in larger oligomers consisting of a double-layer protofibril (a fibril model). The all-atom RMSD is stabilized at around 0.5 nm during the last 20 ns. The increase in RMSD results mostly from the bending of the C-terminal β-sheet spanning residues 39VVIA42. DMFs Binding to Aβ Protofibrils through Both Slow and Fast Binding Processes. We have performed 30 independent MD runs starting from the three different initial states (I, II, III) shown in Figure 1b (10 runs for each initial state). All MD runs use different initial velocity distributions. We monitor the binding status of a DMF molecule to the Aβ protofibril by calculating their minimum distance (dmin). Figure 2 shows the time evolution of the minimum distance between a DMF molecule and the Aβ protofibril. For each initial state (I, II, or III), three representative MD trajectories (denoted as 1, 2, and 3) are presented. The DMF molecule is initially placed 2 nm away from the protofibril. After the MD simulations are initiated, dmin starts to decrease or increase, depending on the initial velocity distributions. Both fast and slow binding processes are observed. The fast binding process is seen in most of the MD runs. For example, in the MD runs of I-1, I-3, II-1, II-2, II-3, III-1, III-2, and III-3, we see that dmin is reduced to 0.35 nm within 25 ns and does not change with the increase of simulation time, indicating that the DMF tightly binds to the Aβ protofibril. A slow binding process is observed in MD run I2, in which dmin reaches 0.35 nm after 50 ns. It is noted that in some of the MD runs, we see that after the DMF molecule gets close to and binds to the protofibril, it stays at the same position during the remaining period of the MD simulation (MD runs I-3, II-2, and III-3 in Figure 2). However, in most of the MD runs, the DMF molecule approaches to the protofibril first but leaves it soon and then gets close to it again. After one or more times of such tries, the DMF molecule finds a suitable position and stays there in the remaining period of MD simulations (see the blue and red curves in Figure 2a, all of the curves in Figure 2b, and the red curve in Figure 2c). We usually stop a MD run if we see that the DMF molecule stays at one binding site for at least 5 ns. To examine the stability of the bound state, we extend three MD runs (I-1, II-1, and III-1) respectively to 90, 70, and 80 ns, in which the DMF starts to bind to the Aβ protofibril at 23.5, 7.5, and 20 ns, respectively. By visual inspection, we find that the DMF molecule retains at the same position during the extended period of simulation; therefore, all of the results presented below are based on the data generated in the 5 ns after binding. The analysis on the Aβ 6736

dx.doi.org/10.1021/jp503458w | J. Phys. Chem. B 2014, 118, 6733−6741

The Journal of Physical Chemistry B

Article

binding energy is positive, indicating that water is unfavorable for the DMF binding. DMFs Binding to Aβ Protofibrils through Two Different Binding Mechanisms. DMF Molecules Binding to the CHC Region via Both Hydrophobic and π-Stacking Interactions. From Table1, we see that three out of the seven binding sites are located around the aromatic residues. There are four aromatic residues (F4, Y10, F19, and F20) in the fulllength Aβ42, and they are all located in the N-terminal β-sheet region. Visual inspection shows that these residues are in close contact with the DMF molecule, especially residues F19−F20 in the CHC region. To further examine the binding strength of the DMF molecule with these residues, we calculate the residue-based binding energy using the method described in the Materials and Methods section, where the ΔEgB term is not considered. We present in Figure 4a−d the residue-based binding energy for the DMF bound to the six different sites 4FRH6, 8SGY10, 13HH14, and CHC sites, the turn site, and the C-terminal β-sheet site. For comparison, the residue-based binding energy at the N-terminal nonaromatic residues site is given in Figure S4 (Supporting Information). Although the binding energy is similar to that at the 4FRH6 site, the contribution of F4 is much smaller. The F and Y residues show lower binding energy with the DMF than their adjacent residues at the same site. The binding energy of the aromatic residues F and Y with the DMF molecule is also lower than that for H13 and H14 residues, indicating that aromatic rings have a stronger interaction with the DMF than with the imidazole rings. At the CHC site, the binding energies of F19 and F20 with the DMF are almost the same and are the lowest (with respect to their adjacent residues at the CHC site), although the side chains of these two residues point to two opposite directions. Visual inspection shows that the DMF molecule binds mostly on the elongation surface (see Figures 3 and S3, Supporting Information), having atomic contacts with the side chains of both F19 and F20. The hydrophobic residues I32 and L34 also interact with the DMF molecule but are weaker than that for F19 and F20 (Figure 4d). Besides the hydrophobic and aromatic stacking interactions, the concave on the elongation surface also facilitates the DMF binding (see Figure 1b). It has been reported that the concaves on the surface of a protein are likely to be the potential binding sites for fullerenes49 or other proteins.50 It is expected that the fullerene molecules may also bind other amyloid peptides if a proper size of hydrophobic grooves exists on the surface of the amyloid peptide monomers, oligomers, or protofibrils. It has been demonstrated that the CHC region plays an important role on the fibrillation of full-length Aβ.51 The binding of DMF molecules to this region on the elongation surface of the Aβ protofibril observed in our MD simulations would block backbone amide sites for further monomer or fibril-competent oligomer addition and retards the elongation process. It has been demonstrated that π-stacking interaction plays an important role in the self-assembly of amyloid fibrils.52 Three different π-stacking orientations have been suggested for aromatic rings in proteins, parallel, T-shaped, and herringbone (around 50°).52,53 To investigate the detailed interaction between the hexatomic carbon ring of the DMF molecule and the aromatic ring of residues 19FF20 in the CHC region, we plot in Figure 5a, d, and g the PDF of the centroid distance between the Phe ring and its closest carbon ring of DMF. There are multiple peaks for their centroid distance in the range of 0−3 nm as there are six CHC regions (six Aβ42 strands) in

protofibril surface. The DMF molecule has a preference to bind to the longitudinal elongation surface along the fibril axis, the lateral β1, and the β2 surfaces rather than the N-terminal surface. We use different colors of spheres to represent the DMF molecules that bind to different sites (see the legend of Figure S3 (Supporting Information) for more details) in all of the 30 MD runs (i.e., there are 30 spheres). We classify the binding sites according to the residue types and spatial configurations, as shown in Table 1. These binding sites are named as follows: N-terminal nonaromatic residues, 4FRH6, 8SGY10, 13HH14, CHC (17LVFFA21), turn site (residues 27NKGAI31), and C-terminal β-sheet site spanning residues 31IIGLMVGGVVI41. The binding sites found only once are classified into “others”. In order to show the binding sites clearly, we present in Figure 3 only one DMF molecule at one binding site.

Figure 3. Binding sites from all of the 30 MD simulations. The DMF molecule is represented as a much smaller sphere at its centroid without the modified group. Multiple binding sites are identified, with three primary binding sites residing at the CHC region, the turn region, and the C-terminal β-sheet region. We use the same colors to represent the residue types as those in Figure 1b.

After identifying these binding sites, we estimate the corresponding binding energy using the MM/GBSA method. It is still a challenge to calculate the accurate binding free energy in solvent. The MM/GBSA method provides a good approximation to consider the solvent effect. The calculated binding energy for each binding site is given in Table 1. It can be seen that the binding energy ΔGbinding at different binding sites varies from about −19 to −35 kcal/mol. The ΔGbinding values at CHC, turn, and C-terminal sites are, respectively, −32.2 ± 4.8, −34.5 ± 5.8, and −34.9 ± 4.3 kcal/mol, much lower than those at other sites, indicating that these three sites are the most favorable ones. The CHC site has also been reported in previous experimental and computational studies,18,25 while the turn and the C-terminal sites are new binding sites found in this study (see below for more detailed discussion). The binding energy components in Table 1 show that, in comparison with other energy terms, the vdW interaction (ΔEvdW) makes a dominant contribution to the total binding energy. The electrostatic interaction between DMF and the Aβ protofibril is relatively weak as there are only four charged atoms with weak polarity on the DMF molecule. Due to the strong hydrophobicity of the DMF molecule, the contribution of the solvent effect (ΔGgB + ΔGSurf) to the total 6737

dx.doi.org/10.1021/jp503458w | J. Phys. Chem. B 2014, 118, 6733−6741

The Journal of Physical Chemistry B

Article

Figure 4. Residue-based binding energy of DMF at six binding sites. The DMF molecule has lower binding energy with aromatic residues (F4, Y10, F19, F20) than that with its adjacent residues. A strong binding affinity is also seen with residue H13-H14, but they are weaker than that with aromatic residues. Note that the EgB term is not taken into account due to the limitation of the algorithm.

the protofibril. The first peak (from left) is centered at (a) 0.40, (d) 0.52, and (g) 0.50 nm, corresponding to the π-stacking interaction. To examine the favorable orientation between the two rings, we plot in Figure 5b, e, and h the PDF of the interplanar ring angle at a centroid distance of 0.3 ≤ d ≤ 0.6 nm (corresponding to the first peak). A single peak at 25° is observed in Figure 5b, indicating that the two rings have a strong preference to be parallel-aligned. PDF peaks centered at around 85° and 55° are also observed (Figure 5e and h), corresponding to respectively the T-shaped and herringbone ring alignments. Representative snapshots (from cluster analysis) showing the three primary ring−ring alignments are given in Figure 5c, f, and I with the side chains of F19 and F20 in bond representation. Similar analyses have also been performed on the aromatic residues at the 4FRH6 and 8SGY10 sites, and we obtain similar results (data not shown). These results are consistent with our previous replica-exchange MD simulation study on the inhibition of Aβ(16−22) aggregation by CNTs, where the parallel and Tshaped orientations between the carbon rings of the CNT and the aromatic rings of Phe residues54 were also observed. DMF Molecules Binding to the Turn Site (27NKGAI31) and the C-Terminal β-Sheet Site (31IIGLMVGGVVI41) via Both Curvature Matching and Hydrophobic Interactions. The turn

binding site involving residues 27NKGAI31 is close to the turn region (23DVGSNK28) of the Aβ protofibril. The binding energy of DMF at this site is about ∼34 kcal/mol, slightly lower than that at the CHC site (−32 kcal/mol). This binding site was not identified in a previous computational study by the docking method.25 To have an intuitive picture of this binding site, we present in Figure 6a the atom details of the Aβ42 peptide in the protofibril with a DMF molecule binding to the turn site. It can be seen that the DMF molecule has a preference to bind to the β2 surface rather than the elongation surface. There is a groove in the 27NKGAI31 region as the side chains of K28 and A30 are in the inner side of the β2-sheet and G29 has the smallest side chain. At the same time, residue I31 has strong hydrophobicity, which is favorable for DMF to bind. In order to examine whether the DMF molecule fits the groove on the surface of the Aβ protofibril, we calculate the surface curvature of the Aβ protofibril without DMF using the Surface Racer software.47 For the Aβ protofibril conformations obtained from the two control MD simulations, the calculated average curvature of each atom is illustrated in Figure 6b. We see a curvature-matched groove (blue) for DMF at the turn binding site. For the C-terminal binding site (31IIGLMVGGVVI41), on one hand, residues G33, G37, and G38 have the smallest side chains, forming a surface groove that fits the DMF 6738

dx.doi.org/10.1021/jp503458w | J. Phys. Chem. B 2014, 118, 6733−6741

The Journal of Physical Chemistry B

Article

Figure 5. Analysis of aromatic stacking interactions between the carbon ring of DMF and the aromatic rings of F19−F20 at the CHC binding site. The PDF of the centroid distance between the Phe ring and the closest carbon ring of DMF, with the first peak centered at (a) 0.4, (d) 0.52, and (g) 0.5 nm. The corresponding PDFs of the interplanar ring angle between the Phe ring and DMF carbon ring with a centroid distance of 0.3 ≤ d ≤ 0.6 nm are given in (b), (e), and (h), respectively. The first PDF peak of the interplanar ring angle in (b), (e), and (h) indicates parallel-aligned (c), Tshaped (f), and herringbone (h) aromatic stacking. The side chains of F19 and F20 in snapshots (c), (f), and (i) are shown in cyan.

the DMF−K28 interaction impacts the stability of the D23− K28 salt bridge, we plot in Figure 7 the PDF of the distance

Figure 7. PDF of the distance between D23 and K28 in the first/sixth peptide with DMF binding to the turn region, CHC region, and Cterminal β-sheet region. For comparison, the corresponding PDF of the D23−K28 distance for the Aβ hexamer without DMF (black curve) is also shown.

between D23 and K28 in the first and the sixth peptides. For comparison, we also present in Figure 7 the PDF curves of the D23−K28 distance without DMF and with DMF binding to the CHC and the C-terminal sites. It can be seen that there exists only one peak centered at 0.35 nm in the absence of DMF and with DMF binding to the CHC and C-terminal sites, indicating that the D23−K28 salt bridges are well preserved. However, when the DMF molecule binds to the turn site, the sharp PDF peak at 0.35 nm becomes much smaller, and a broad peak at 1.25 nm appears. The appearance of the broad peak at 1.25 nm implies that the binding of DMF to the turn site disrupts the salt bridge. It has been reported that the D23−K28 salt bridge plays an important role on Aβ fibrillation.55 The destabilization effect of DMF on the D23−K28 salt bridge observed here would explain the inhibition effect of DMF on the Aβ fibrillation reported previously by in vitro experiment.27 It is noted that the DMF/Aβ molar ratio is 1:6 in our simulated

Figure 6. (a) The atom details of Aβ42 with DMF binding to the turn site. (b) The curvature representation of the molecular surface of the Aβ hexamer averaged over all conformations. Blue color indicates curvature-matched sites for DMF binding, while red is for mismatched sites.

well (see Figure 6b). On the other hand, 31II32, 34LM35, and 39VVI41 are hydrophobic residues, facilitating DMF to bind. These two factors lead to a low binding energy (−35 kcal/mol) of DMF at the C-terminal β-sheet site. Upon binding to the turn site of the Aβ protofibril, the DMF molecule has atomic contacts with K28. To examine whether 6739

dx.doi.org/10.1021/jp503458w | J. Phys. Chem. B 2014, 118, 6733−6741

The Journal of Physical Chemistry B

Article

(2) Selkoe, D. J. Alzheimer’s Disease: Genes, Proteins, and Therapy. Physiol. Rev. 2001, 81, 741−766. (3) 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. (4) Selkoe, D. J. Translating Cell Biology into Therapeutic Advances in Alzheimer’s Disease. Nature 1999, 399, A23−A31. (5) Pitschke, M.; Prior, R.; Haupt, M.; Riesner, D. Detection of Single Amyloid β-Protein Aggregates in the Cerebrospinal Fluid of Alzheimer’s Patients by Fluorescence Correlation Spectroscopy. Nat. Med. 1998, 4, 832−834. (6) Serpell, L. C.; Sunde, M.; Benson, M. D.; Tennent, G. A.; Pepys, M. B.; Fraser, P. E. The Protofilament Substructure of Amyloid Fibrils. J. Mol. Biol. 2000, 300, 1033−1039. (7) Luhrs, T.; Ritter, C.; Adrian, M.; Riek-Loher, D.; Bohrmann, B.; Doeli, H.; Schubert, D.; Riek, R. 3D Structure of Alzheimer’s Amyloidβ(1−42) Fibrils. Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 17342−17347. (8) Tycko, R. Solid-State NMR Studies of Amyloid Fibril Structure. Annu. Rev. Phys. Chem. 2011, 62, 279−299. (9) Walsh, D. M.; Selkoe, D. J. Aβ Oligomers  A Decade of Discovery. J. Neurochem. 2007, 101, 1172−1184. (10) Lomakin, A.; Teplow, D. B.; Kirschner, D. A.; Benedek, G. B. Kinetic Theory of Fibrillogenesis of Amyloid β-Protein. Proc. Natl. Acad. Sci. U.S.A. 1997, 94, 7942−7947. (11) Straub, J. E.; Thirumalai, D. Principles Governing Oligomer Formation in Amyloidogenic Peptides. Curr. Opin. Struc. Biol. 2010, 20, 187−195. (12) Straub, J. E.; Thirumalai, D. Toward a Molecular Theory of Early and Late Events in Monomer to Amyloid Fibril Formation. Annu. Rev. Phys. Chem. 2011, 62, 437−463. (13) Miller, Y.; Ma, B.; Nussinov, R. Zinc Ions Promote Alzheimer Aβ Aggregation via Population Shift of Polymorphic States. Proc. Natl. Acad. Sci. U.S.A. 2010, 107, 9490−9495. (14) DeToma, A. S.; Salamekh, S.; Ramamoorthy, A.; Lim, M. H. Misfolded Proteins in Alzheimer’s Disease and Type II Diabetes. Chem. Soc. Rev. 2012, 41, 608−621. (15) Cabaleiro-Lago, C.; Quinlan-Pluck, F.; Lynch, I.; Dawson, K. A.; Linse, S. Dual Effect of Amino Modified Polystyrene Nanoparticles on Amyloid-β Protein Fibrillation. ACS. Chem. Neurosci. 2010, 1, 279− 287. (16) Cabaleiro-Lago, C.; Quinlan-Pluck, F.; Lynch, I.; Lindman, S.; Minogue, A. M.; Thulin, E.; Walsh, D. M.; Dawson, K. A.; Linse, S. Inhibition of Amyloid-β Protein Fibrillation by Polymeric Nanoparticles. J. Am. Chem. Soc. 2008, 130, 15437−15443. (17) Podolski, I. Y.; Podlubnaya, Z. A.; Kosenko, E. A.; Mugantseva, E. A.; Makarova, E. G.; Marsagishvili, L. G.; Shpagina, M. D.; Kaminsky, Y. G.; Andrievsky, G. V.; Klochkov, V. K. Effects of Hydrated Forms of C60 Fullerene on Amyloid β-Peptide Fibrillization In Vitro and Performance of the Cognitive Task. J. Nanosci. Nanotechnol. 2007, 7, 1479−1485. (18) Kim, J. E.; Lee, M. Fullerene Inhibits β-Amyloid Peptide Aggregation. Biochem. Biophys. Res. Commun. 2003, 303, 576−579. (19) Huang, H. M.; Ou, H. C.; Hsieh, S. J.; Chiang, L. Y. Blockage of Amyloid β Peptide-Induced Cytosolic Free Calcium by Fullerenol-1, Carboxylate C60 in PC12 Cells. Life Sci. 2000, 66, 1525−1533. (20) Linse, S.; Cabaleiro-Lago, C.; Xue, W. F.; Lynch, I.; Lindman, S.; Thulin, E.; Radford, S. E.; Dawson, K. A. Nucleation of Protein Fibrillation by Nanoparticles. Proc. Natl. Acad. Sci. U.S.A. 2007, 104, 8691−8696. (21) Dugan, L. L.; Turetsky, D. M.; Du, C.; Lobner, D.; Wheeler, M.; Almli, C. R.; Shen, C. K. F.; Luh, T. Y.; Choi, D. W.; Lin, T. S. Carboxyfullerenes as Neuroprotective Agents. Proc. Natl. Acad. Sci. U.S.A. 1997, 94, 9434−9439. (22) Lee, C. M.; Huang, S. T.; Huang, S. H.; Lin, H. W.; Tsai, H. P.; Wu, J. Y.; Lin, C. M.; Chen, C. T. C60 Fullerene−Pentoxifylline Dyad Nanoparticles Enhance Autophagy to Avoid Cytotoxic Effects Caused by the β-Amyloid Peptide. Nanomedicine: Nanotechnology, Biology and Medicine 2011, 7, 107−114.

system in order to save simulation time; therefore, the destabilization effect on the Aβ protofibril is limitted. It is expected that the salt bridge disruption effect would be enhanced when the DMF/Aβ molar ratio is increased.27

4. CONCLUSIONS We have studied the binding process and the detailed interaction of a water-soluble fullerene derivative (DMF) with a fibrillar Aβ42 hexamer (a protofibril model) using the all-atom MD simulations. By conducting extensive multiple MD simulations initiated from three different states with the DMF molecule placed 2.0 nm away from the protofibril, we have identified three primary binding sites residing in the CHC region (17LVFFA21), the turn region (27NKGAI31), and the C-terminal β-sheet region (31IIGLMVGGVVI41). The binding of DMF to these sites is driven by different mechanisms. The πstacking plays a crucial role in binding affinity, which leads to a series of binding sites such as CHC, 4FRH6, and 8SGY10. As an important hydrophobic segment for aggregation, the CHC sites also reported in other work. In addition, the curvature fitting mechanism was also found to be an important feature for fullerene binding with amyloid peptides. The sites located in turn region were first observed in this work. Furthermore, DMF in this site destabilized the D23−K28 salt bridge, which was considered to be important for the amyloid fibril. Our results provide a better understanding of the binding mechanism and fibril inhibition effect of fullerene derivatives, which may be helpful in drug design of AD therapeutics.

■

ASSOCIATED CONTENT

S Supporting Information *

Two tables showing the force field of the DMF molecule and a summary of the binding sites. Four figures showing the conformational analysis of the isolated fibrillar Aβ42 hexamer, the time evolution of the minimum distance (dmin) between the dimethoxymethoxyane group of DMF and the Aβ hexamer after DMF binding, the binding sites from all of the 30 MD simulations, and the residue-based binding energy of DMF at the N-terminal nonaromatic residues site. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*E-mails: [email protected]. Tel: 86-21-55665231. Author Contributions †

X.Z. and W.X. contributed equally to this work.

Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS We thank Dr. Ruhong Zhou and Dr. Buyong Ma for helpful discussion. This work was supported by the National Natural Science Foundation of China (Grant No.: 91227102 and 11274075). X.Z. acknowledges the Hui-Chun Chin and TsungDao Lee Chinese Undergraduate Research Endowment. Simulations were performed at the Shanghai Supercomputing Center and the National High Performance Computing Center of Fudan University.

■

REFERENCES

(1) Chiti, F.; Dobson, C. M. Protein Misfolding, Functional Amyloid, and Human Disease. Annu. Rev. Biochem. 2006, 75, 333−366. 6740

dx.doi.org/10.1021/jp503458w | J. Phys. Chem. B 2014, 118, 6733−6741

The Journal of Physical Chemistry B

Article

(23) Grebowski, J.; Kazmierska, P.; Krokosz, A. Fullerenols as a New Therapeutic Approach in Nanomedicine. BioMed Res. Int. 2013, 2013, 751913. (24) Dugan, L. L.; Lovett, E. G.; Quick, K. L.; Lotharius, J.; Lin, T. T.; O’Malley, K. L. Fullerene-Based Antioxidants and Neurodegenerative Disorders. Parkinsonism Relat. Disord. 2001, 7, 243−246. (25) Andujar, S. A.; Lugli, F.; Hofinger, S.; Enriz, R. D.; Zerbetto, F. Amyloid-β Fibril Disruption by C60 Molecular Guidance for Rational Drug Design. Phys. Chem. Chem. Phys. 2012, 14, 8599−8607. (26) Wu, C.; Scott, J.; Shea, J. E. Binding of Congo Red to Amyloid Protofibrils of the Alzheimer Aβ9−40 Peptide Probed by Molecular Dynamics Simulations. Biophys. J. 2012, 103, 550−557. (27) Kim, J. E.; Lee, M. Fullerene Inhibits β-Amyloid Peptide Aggregation. Biochem. Biophys. Res. Commun. 2003, 303, 576−579. (28) Hezaveh, S.; Samanta, S.; Milano, G.; Roccatano, D. Structure and Dynamics of 1,2-Dimethoxyethane and 1,2-Dimethoxypropane in Aqueous and Non-Aqueous Solutions: A Molecular Dynamics Study. J. Chem. Phys. 2011, 135, 164501. (29) Ma, B. Y.; Nussinov, R. Simulations as Analytical Tools to Understand Protein Aggregation and Predict Amyloid Conformation. Curr. Opin. Chem. Biol. 2006, 10, 445−452. (30) Lu, J. X.; Qiang, W.; Yau, W. M.; Schwieters, C. D.; Meredith, S. C.; Tycko, R. Molecular Structure of β-Amyloid Fibrils in Alzheimer’s Disease Brain Tissue. Cell. 2013, 154, 1257−1268. (31) Hess, B.; Kutzner, C.; van der Spoel, D.; Lindahl, E. GROMACS 4: Algorithms for Highly Efficient, Load-Balanced, and Scalable Molecular Simulation. J. Chem. Theory Comput. 2008, 4, 435−447. (32) van Gunsteren, W. F.; Billeter, S.; Eising, A.; Hünenberger, P. H.; Krüger, P.; Mark, A. E.; Scott, W.; Tironi, I. G. Biomolecular Simulation: The GROMOS96 Manual and User Guide; Vdf Hochschulverlag AG, an der ETH Zurich, Switzerland, 1996. (33) Berendsen, H.; Postma, J.; Van Gunsteren, W.; Hermans, J. Interaction Models for Water in Relation to Protein Hydration. Intermol. Forces 1981, 11, 331−342. (34) Hess, B.; Bekker, H.; Berendsen, H. J.; Fraaije, J. G. LINCS: A Linear Constraint Solver for Molecular Simulations. J. Comput. Chem. 1997, 18, 1463−1472. (35) Miyamoto, S.; Kollman, P. A. Settle: An Analytical Version of the SHAKE and RATTLE Algorithm for Rigid Water Models. J. Comput. Chem. 1992, 13, 952−962. (36) Darden, T.; York, D.; Pedersoen, L. Particle Mesh Eward: An N log(N) Method for Ewald Sums in Large Systems. J. Phys. Chem. B 1993, 98, 10089. (37) Bussi, G.; Donadio, D.; Parrinello, M. Canonical Sampling through Velocity Rescaling. J. Chem. Phys. 2007, 126, 014101. (38) Parrinello, M.; Rahman, A. Polymorphic Transitions in SingleCrystals  A New Molecular Dynamics Method. J. Appl. Phys. 1981, 52, 7182−7190. (39) Humphrey, W.; Dalke, A.; Schulten, K. VMD: Visual Molecular Dynamics. J. Mol. Graphics 1996, 14, 33−38. (40) Case, D. A.; Darden, T. A.; Cheatham, T. E. III; Simmerling, C. L.; Wang, J.; Duke, R. E.; Luo, R.; Walker, R. C.; Zhang, W.; Merz, K. M. AMBER 11; University of California: San Francisco, CA, 2010. (41) Onufriev, A.; Bashford, D.; Case, D. A. Exploring Protein Native States and Large-Scale Conformational Changes with a Modified Generalized Born Model. Proteins 2004, 55, 383−394. (42) Onufriev, A.; Bashford, D.; Case, D. A. Modification of the Generalized Born Model Suitable for Macromolecules. J. Phys. Chem. B 2000, 104, 3712−3720. (43) Wang, Q.; Yu, X.; Patal, K.; Hu, R.; Chuang, S.; Zhang, G.; Zheng, J. Tanshinones Inhibit Amyloid Aggregation by Amyloid-β Peptide, Disaggregate Amyloid Fibrils, and Protect Cultured Cells. ACS. Chem. Neurosci. 2013, 4, 1004−1015. (44) Zhang, T.; Zhang, J.; Derreumaux, P.; Mu, Y. Molecular Mechanism of the Inhibition of EGCG on the Alzheimer Aβ1-42 Dimer. J. Phys. Chem. B 2013, 117, 3993−4002. (45) Berhanu, W. M.; Hansmann, U. H. E. The Stability of Cylindrin β-Barrel Amyloid Oligomer ModelsA Molecular Dynamics Study. Proteins 2013, 81, 1542−1555.

(46) Kopitz, H.; Zivkovic, A.; Engels, J. W.; Gohlke, H. Determinants of the Unexpected Stability of RNA Fluorobenzene Self Pairs. ChemBioChem 2008, 9, 2619−2622. (47) Tsodikov, O. V.; Record, M. T.; Sergeev, Y. V. Novel Computer Program for Fast Exact Calculation of Accessible and Molecular Surface Areas and Average Surface Curvature. J. Comput. Chem. 2002, 23, 600−609. (48) Horn, A. H. C.; Sticht, H. Amyloid-β42 Oligomer Structures from Fibrils: A Systematic Molecular Dynamics Study. J. Phys. Chem. B 2010, 114, 2219−2226. (49) Benyamini, H. Interaction of C60-Fullerene and Carboxyfullerene with Proteins: Docking and Binding Site Alignment. Bioconjugate Chem. 2006, 17, 378−386. (50) Ma, B.; Elkayam, T.; Wolfson, H.; Nussinov, R. Protein−Protein Interactions: Structurally Conserved Residues Distinguish between Binding Sites and Exposed Protein Surfaces. Proc. Natl. Acad. Sci. U.S.A. 2003, 100, 5772−5777. (51) Tjernberg, L. O.; Näslund, J.; Lindqvist, F.; Johansson, J.; Karlström, A. R.; Thyberg, J.; Terenius, L.; Nordstedt, C. Arrest of βAmyloid Fibril Formation by a Pentapeptide Ligand. J. Biol. Chem. 1996, 271, 8545−8548. (52) Gazit, E. A Possible Role For Π-Stacking in the Self-Assembly of Amyloid Fibrils. FASEB J. 2002, 16, 77−83. (53) Sun, S.; Bernstein, E. Aromatic van der Waals Clusters: Structure and Nonrigidity. J. Phys. Chem. 1996, 100, 13348−13366. (54) Li, H.; Luo, Y.; Derreumaux, P.; Wei, G. Carbon Nanotube Inhibits the Formation of β-Sheet-Rich Oligomers of the Alzheimer’s Amyloid-β(16−22) Peptide. Biophys. J. 2011, 101, 2267−2276. (55) Reddy, G.; Straub, J. E.; Thirumalai, D. Influence of Preformed Asp23-Lys28 Salt Bridge on the Conformational Fluctuations of Monomers and Dimers of Aβ Peptides with Implications for Rates of Fibril Formation. J. Phys. Chem. B 2009, 113, 1162−1172.

6741

dx.doi.org/10.1021/jp503458w | J. Phys. Chem. B 2014, 118, 6733−6741