Effects of Ligands on Unfolding of the Amyloid β-Peptide Central Helix: Mechanistic Insights from Molecular Dynamics Simulations

Polymerization of the amyloid β-peptide (Aβ), a process which requires that the helical structure of Aβ unfolds beforehand, is suspected to cause neurodegeneration in Alzheimer's disease. According to recent experimental studies, stabilization of the Aβ central helix counteracts Aβ polymerization into toxic assemblies. The effects of two ligands (Dec-DETA and Pep1b), which were designed to bind to and stabilize the Aβ central helix, on unfolding of the Aβ central helix were investigated by molecular dynamics simulations. It was quantitatively demonstrated that the stability of the Aβ central helix is increased by both ligands, and more effectively by Pep1b than by Dec-DETA. In addition, it was shown that Dec-DETA forms parallel conformations with β-strand-like Aβ, whereas Pep1b does not and instead tends to bend unwound Aβ. The molecular dynamics results correlate well with previous experiments for these ligands, which suggest that the simulation method should be useful in predicting the effectiveness of novel ligands in stabilizing the Aβ central helix. Detailed Aβ structural changes upon loss of helicity in the presence of the ligands are also revealed, which gives further insight into which ligand may lead to which path subsequent to unwinding of the Aβ central helix.


Introduction
Alzheimer's disease (AD) is one of the most common neurodegenerative disorders in aging people. According to the amyloid cascade hypothesis [1,2,3], accumulation of the amyloid b-peptide (Ab) in the brain is the primary influence driving AD pathogenesis. Originally insoluble fibrils and plaques composed of Ab were suspected to cause AD [1,2], but currently prefibrillar aggregates including soluble oligomers composed of Ab are also considered to be the cause of AD [3]. Ab is produced mainly as a 40-or 42-residue peptide by proteolysis of an integral membrane protein, the amyloid precursor protein (APP). Nuclear magnetic resonance (NMR) data showed that Ab(1-40) adopts a folded structure including two a-helical regions (residues 15-24 and 29-35) in water/sodium dodecyl sulfate (SDS) micelles which provide a water-membrane interface mimicking environment [4,5], and that Ab(1-42) adopts an unfolded structure including two bstrands (residues 17-21 and 31-36) in aqueous solution [6]. Using NMR it has also been shown that an Ab(1-42) fibril is a b-sheet composed of two b-strands (residues 18-26 and 31-42) [7]. These structural data indicate that, once Ab departs from the membrane to the extracellular fluid, its a-helical regions unfold to elongated or b-strand-like forms, and that the b-strands of Ab enable formation of b-sheets of fibrils and prefibrillar aggregates.
A wide range of molecules including small compounds and synthetic peptide derivatives have been identified as anti-amyloid agents [8]. Most of these molecules are predicted to bind to elongated or b-strand-like Ab and to inhibit b-sheet extension, and thus they are expected to prevent Ab polymerization. However, this strategy may be problematic in that it will favor formation of prefibrillar aggregates such as Ab oligomers which are cytotoxic [9], and that some of the ligands may act as aggregators [10]. Alternative strategies to develop anti-amyloid agents are needed to overcome these problems. Earlier steps in amyloidogenesis before emergence of b-strand-like Ab should be targeted to pursue alternative strategies. The emergence of b-strand-like Ab can be inhibited by trapping Ab in a state similar to its native structure in membrane embedded APP.
Recent experimental studies [11,12] demonstrated that trapping Ab in a state similar to its native structure by stabilizing the Ab central helix (residues [15][16][17][18][19][20][21][22][23][24] is an effective strategy to reduce Ab polymerization and Ab toxicity. Two different classes of ligands were designed to bind and stabilize the Ab central helix, and it was shown that in the presence of either ligand, Ab helical content was increased, the amount of Ab fibrils was reduced, Ab toxicity to PC12 cells in culture and to hippocampal slice preparations was reduced, and the lifespan of Drosophila model was prolonged [12]. Although many effects of the two ligands (Dec-DETA and Pep1b) are similar, there are also different effects on polymerization. That is, thicker-than-normal Ab fibrils were detected in the presence of Dec-DETA, and shorter-than-normal Ab fibrils were detected in the presence of Pep1b, though both ligands substantially reduced the amount of Ab fibrils. The reason for this was not clarified in the experimental study. We suspect that there are differences in behavior toward Ab between the two ligands.
In order to rationally design new compounds that more effectively stabilize the Ab central helix and reduce Ab polymerization into toxic assemblies, detailed molecular mechanisms that underlie unfolding and stabilization of the Ab central helix should be elucidated. Elucidation of such detailed molecular mechanisms, which are difficult to analyze by using only experimental methods, is possible by taking advantage of computational methods like molecular dynamics (MD). The unfolding process of the Ab helix has attracted much attention and has been studied by MD simulations [13,14,15,16,17]. However, effects of ligands on the unfolding process of the Ab helix have not been fully investigated and detailed molecular mechanisms for the Ab helix stabilization by ligands have not been uncovered yet, though short MD simulations indicated that the designed ligands stabilize the a-helical conformation of Ab(13-26) [12].
Structures of the two ligand-peptide complexes were manually built using the Insight II program to satisfy the ligand-peptide contacts that were intended in their design: The Dec-DETA complex was designed for electrostatic interaction with E22 and D23 via the two basic functional groups and for van der Waals interaction with L17, V18, and A21 via the hydrocarbon tail ( Fig. 1); similarly the Pep1b complex was built for electrostatic interaction with E22 and D23 via the two basic functional groups and with H13 and K16 via the two acidic functional groups, and for van der Waals interaction with F20 via the indole group (Fig. 1).

MD Simulations
All calculations were carried out using the CHARMM22/ CMAP force field [26,27,28] with the CHARMM program [29,30]. The force field parameters for the ligands (Table S1) were picked from the CHARMM22 force field parameters for proteins, since the ligands were designed basically using amino acid moieties. The SHAKE [31] algorithm was applied to fix all covalent bonds containing a hydrogen atom allowing a 2 fs timestep to be used in the integration of Newton's equations. The nonbonded (van der Waals and Coulomb) interaction energies and forces were smoothly shifted to zero at 12 Å using the atom-based force-shift method [32,33], and the nonbonded list was constructed with a cutoff of 16 Å and was updated every time any atom moved by more than 2 Å since the last update. Before MD simulations were carried out, structures of the solvated systems were optimized by 500 steps of steepest descent energy minimization with a harmonic restraint of 20 kcal/mol/Å 2 on Ab followed by 1500 steps of adopted basis Newton-Raphson energy minimization without a harmonic restraint on Ab (Fig. 1). After the systems were heated up to 360 K gradually for 50 ps, ten independent 20 ns MD simulations at 360 K with different initial velocity assignments were carried out for each system to increase sampling [34]. The MD simulations were performed for the optimized systems under periodic boundary conditions at a constant pressure (1 atm) using the Langevin piston method [35] with piston mass 400 amu, collision frequency 20 ps 21 and bath temperature (360 K). The average temperature was checked every 4 ps, and was found to remain within 5 K of the target temperature after the heating MD run. Fast table lookup routines for non-bonded interactions [36] were used to increase speed of the MD simulations. During the MD simulations, no harmonic restraints were imposed on any molecule in the systems, and coordinates were saved every 1 ps.
In our previous study [17], we showed that the Ab central helix completely unfolded at 360 K in 20 ns MD simulations, though it did not unfold at the lower temperatures (300 and 330 K). Therefore, the simulations for each system were performed at 360 K to accelerate dynamics of Ab. Additionally, one control 20 ns MD simulation for each system was performed at 310 K with the methods used for the MD simulations at 360 K.

Analyses
All analyses were carried out for the trajectories obtained by the MD simulations at 360 K, except as otherwise stated. The data of every 10 ps of the trajectories after the heating time of the MD simulations were used for the analyses. Visualization of the structural change of the Ab and Ab-ligand complex models during MD simulations was carried out by using the visual molecular dynamics (VMD) software (version 1.8.6) [37].
To examine how each ligand interacted with Ab during the simulations, the probability of the contact between the center of geometry of sidechain heavy atoms of each Ab residue and each ligand heavy atom was calculated, using the criterion distance #6.0 Å . The distance criterion (6.0 Å ) was chosen considering the contact distances measured for the initial energy-minimized structures of the Ab-ligand complexes. A map of the contacts between Ab and Dec-DETA or Pep1b was created using the calculated probabilities.
To determine details of polar interactions between Ab and each ligand, the number of hydrogen bonds (HBs) between Ab and Dec-DETA or Pep1b was calculated, using the criterion acceptor-hydrogen distance #2.4 Å . For this calculation, both of HBs between Ab sidechain atoms and ligand atoms and HBs between Ab backbone atoms and ligand atoms were counted (for each Ab-ligand complex, the number of the latter HBs was less than 10% of that of all the HBs). When at least one HB between Ab and the ligand was observed, the ligand was considered to be bound to Ab.
Additionally, to determine details of nonpolar interactions between Ab and each ligand, the number of C-C and C-N contacts between carbon atoms of the Ab middle nonpolar part (residues 17-21) and nine heavy atoms of the Dec-DETA hydrocarbon tail (C1-C9) or of the Pep1b indole group (C13-C20 and N3) was calculated, using the criterion C-C or C-N distance #5.0 Å . The backbone carbonyl carbon atoms of Ab were not included in this calculation. The distance criterion (5.0 Å ) was chosen considering the radii of carbon (1.8-2.3 Å ), nitrogen (1.9 Å ), and hydrogen (1.3-1.4 Å ) atoms and the C-H and N-H covalent bond lengths (1.0-1.1 Å ) used in the CHARMM22 force field [26]. When at least one contact between the Ab middle nonpolar part and the ligand nonpolar part was observed, the ligand nonpolar part was considered to be in contact with the Ab middle nonpolar part. In this analysis, contacts between the Ab middle nonpolar part and the ligand nonpolar parts (the hydrocarbon tail of Dec-DETA and the indole group of Pep1b) were focused on, because it was shown that, during the simulations, the ligand nonpolar parts were mainly in contact with the Ab middle nonpolar part as they were designed ( Fig. 1).

Effects of the Ligands on Stability of the Ab Central Helix
To examine whether the Ab central helix eventually unfolded by the end of the simulation, the average backbone RMSD of the Ab middle region (15)(16)(17)(18)(19)(20)(21)(22)(23)(24) and the average number of aHBs of the Ab middle region calculated for the last 2 ns of the each 20 ns simulation, where fluctuation of the Ab backbone RMSD is relatively small in every trajectory, were analyzed ( Table 1). The trajectories were classified into three groups: group A (RMSD,2.0 Å , 2#aHB#6), group B (2.0 Å #RMSD,4.0 Å , 1#aHB#4), and group C (RMSD$4.0 Å , aHB<0). By visual inspection, it was ascertained that the Ab central helix maintained its helical conformation during the whole simulations or refolded after partial unfolding by the end of the simulations in the group A trajectories, that it partially unfolded by the end of the simulations in the group B trajectories, and that it completely unfolded by the end of the simulations in the group C trajectories. The helical Ab (group A) is observed in only one trajectory in the absence of a ligand, whereas it is observed in five trajectories in the presence of Dec-DETA and is observed in four trajectories in the presence of Pep1b (Table 1). In contrast, the completely unfolded Ab (group C) is observed in three trajectories in the absence of a ligand, whereas it is observed in only one trajectory in the presence of Dec-DETA and is not observed in any trajectory in the presence of Pep1b (Table 1).
To examine behavior of the Ab middle region during the simulations, the backbone RMSD during the whole simulations ( Fig. 2A) and during the second half of the simulations (Fig. 2B) was calculated. By analyzing the backbone RMSD of the whole simulation of each trajectory, it was found that the Ab helix was relatively stable during the first half of the simulations in five out of ten trajectories even if a ligand was not added to the system. For this reason, the second half of the simulations was used for this analysis. By visual inspection, it was determined that Ab structures with small (RMSD,2.0 Å ), medium (2.0 Å #RMSD,4.0 Å ), and large (RMSD$4.0 Å ) RMSD correspond to helical, moderately unwound, and highly unwound or elongated Ab structures, respectively. Below we refer to these groups as peptide-conformation classes 1, 2, and 3, respectively. Both ligands, particularly Pep1b, increase the population of class 1 and decrease the population of class 3 (Fig. 2). During the second half of the simulations, the relative frequencies of class 1 and 3 in the presence of Dec-DETA are 1.6 and 0.5 times the frequencies for Ab alone. In the presence of Pep1b the corresponding numbers are 2.1 and 0.2. Without a ligand class 3 is more populated than class 1 during the second half of the simulations, a situation which is reversed by both ligands (Fig. 2B).
The number of aHBs in the helix was calculated to further characterize the behavior of the Ab middle region (Fig. 3). The relative frequency of Ab structures with no aHBs is decreased by addition of both ligands, particularly by addition of Pep1b (Fig. 3). This aspect is observed especially in the second half of the simulations (Fig. 3B). The existence of Ab structures with five or six aHBs is increased by addition of both ligands, particularly by addition of Pep1b. During the second half of the simulations, the probability to find at least five aHBs is 1.3 and 1.5 times higher for Ab in the presence of Dec-DETA and Pep1b, respectively, compared to Ab alone.
These results indicate that both addition of Dec-DETA and Pep1b are effective in stabilizing the Ab central helix and that Pep1b is somewhat more effective than Dec-DETA.

Interactions between the Ligands and Ab
To examine whether the ligands were in contact with Ab as they were designed (Fig. 1), the contact maps ( Fig. 4 and 5) were analyzed. All contact probabilities are lower than 0.6, indicating that the ligands sometimes detached from Ab. By visual inspection of the trajectories, we found both Ab and the ligands to be quite flexible and that the ligands sometimes detached from Ab but bound to Ab again. High contact probabilities (0.4#P,0.6) are  observed for contacts between the basic functional groups (N2 and N3) of Dec-DETA and the acidic residues (E22 and D23) of Ab and for contacts between the basic functional groups (N5, N7, and N8) of Pep1b and the acidic residues (E22 and D23) of Ab. Contacts between the acidic functional groups (O1, O2, O4, and O5) of Pep1b and the basic residues (H13 and K16) of Ab occur with medium probabilities (0.2#P,0.3). Contacts between the Dec-DETA hydrocarbon tail (C1-C9) and the Ab middle nonpolar part are distributed from L17 to A21 of Ab, although the probabilities are low (0.1#P,0.2). In contrast, contacts between the Pep1b indole group (C13-C20 and N3) and the Ab middle nonpolar part are localized at F19 and F20 of Ab, with a preference for F20 (0.2#P,0.3). Thus, the contact maps show that the ligands were in contact with Ab as they were designed, even though the ligands sometimes detached from Ab.
Contact maps from simulations of both Ab-ligand complexes at 310 K ( Fig. S1 and S2) show higher probabilities (P$0.6) than at 360 K, and the distribution of contacts in each Ab-ligand complex is more localized at 310 K than at 360 K. This is because the conformations of Ab and the ligands did not change so much and the ligands almost always bound to Ab at 310 K, in contrast to the motions of Ab and the ligands at 360 K. However, the pattern of contacts in each Ab-ligand complex at 310 K is similar to that at 360 K, and the main contacts of each Ab-ligand complex at 310 K are almost the same as those at 360 K. Although motions of the ligands and Ab are enhanced due to the increased temperature, interactions between the ligands and Ab at the relatively high temperature are thus similar to those at the body temperature.
To understand polar interactions between the ligands and Ab, the existence of HBs between the ligands and Ab was analyzed for the three peptide-conformation classes; the frequency of time when the ligands do not form any HBs with Ab regardless of the peptide conformation was also calculated (Fig. 6A). In total, Dec-DETA and Pep1b form at least one HB with Ab for 73% and 91% of the total time, respectively (Fig. 6A). When we consider only the helical class 1 conformations, Pep1b is in polar contact (hydrogen bonding contact) with Ab 1.7 times as often as Dec-DETA (Fig. 6A). The fraction of the occurrence of the polar contacts for each peptide-conformation class ( Table 2) shows that Pep1b binds to the Ab structures in class 1 with higher probability than to the Ab structures in classes 2 and 3, whereas Dec-DETA binds to all three peptide-conformation classes with similar probabilities. Besides, the fraction of the occurrence of the polar contacts for the class 1 conformations is higher for Pep1b than for Dec-DETA (Table 2). These data indicate that Pep1b binds more specifically to helical Ab than Dec-DETA does. Additionally, the Ab structures in class 1 form one more HB on average with Pep1b than with Dec-DETA (Table 3).
In a similar way, we analyzed the existence of nonpolar interactions (C-C and C-N contacts) between the nonpolar groups of the ligands (the hydrocarbon tail of Dec-DETA and the indole The probability (0.0#P,0.6) of the contact between the center of geometry of sidechain heavy atoms of each Ab residue and each Dec-DETA heavy atom is colored (white to blue grids). The probability was calculated using the data obtained from the whole simulations of all ten trajectories. The Ab residues and Dec-DETA atoms corresponding to the X and Y-axis numbers, respectively, are listed below the map. doi:10.1371/journal.pone.0030510.g004 Figure 5. Contact map of the Ab-Pep1b complex. The probability (0.0#P,0.6) of the contact between the center of geometry of sidechain heavy atoms of each Ab residue and each Pep1b heavy atom is colored (white to blue grids). The probability was calculated using the data obtained from the whole simulations of all ten trajectories. The Ab residues and Pep1b atoms corresponding to the X and Y-axis numbers, respectively, are listed below the map. doi:10.1371/journal.pone.0030510.g005 group of Pep1b) and the middle nonpolar part (residues 17-21) of Ab for the three peptide-conformation classes; the frequency of time when the ligands do not have any C-C and C-N contacts with Ab regardless of the peptide conformation was also calculated (Fig. 6B). In total, the nonpolar groups of Dec-DETA and Pep1b have at least one C-C or C-N contact with the middle nonpolar part of Ab for 64% and 69% of the total time, respectively (Fig. 6B). When we consider only the class 1 conformations, Pep1b is in nonpolar contact with Ab 1.4 times as often as Dec-DETA (Fig. 6B). The fraction of the occurrence of the nonpolar contacts for the class 1 conformations is higher for Pep1b than for Dec-DETA (Table 2). These data indicate that the indole group of Pep1b has contacts with the middle nonpolar part of helical Ab more frequently than the hydrocarbon tail of Dec-DETA does.
Additionally, the Ab structures in class 1 have one more C-C or C-N contact on average with Pep1b than with Dec-DETA (Table 3).
To further understand interactions between the ligands and Ab, we also anlyzed the existence of HBs between the ligands and Ab for the three peptide-conformation classes in each individual trajectory (Fig. 7). The intermittent lines for the Ab-Dec-DETA (Fig. 7A) and Ab-Pep1b (Fig. 7B) complexes show that both ligands sometimes detach from Ab and bind again to Ab. Long durations of the ligands in hydrogen bonding contact with the class 1 conformations are more frequent for Ab-Pep1b (Fig. 7B) than for Ab-Dec-DETA (Fig. 7A). This shows that, compared to Dec-DETA, Pep1b binds to the helical conformations of Ab more constantly and is thus more effective in stabilizing the Ab central helix. In contrast, long durations of the ligands in hydrogen bonding contact with the class 3 conformations are more frequent for Ab-Dec-DETA than for Ab-Pep1b, indicating that Dec-DETA binds to the highly unwound or elongated conformations of Ab for longer periods than Pep1b.
In addition, to examine whether Pep1b binds to Ab with both acidic and basic functional groups at the same time during the simulation, we analyzed events when both basic and acidic functional groups of Pep1b form HBs with the sidechains of the acidic and basic residues of Ab, respectively, at the same time (Fig. 7C). All trajectories begin with Ab in conformation class 1  Table 3. Average number of polar and nonpolar contacts between Ab and Dec-DETA or Pep1b for each peptideconformation class a . and Pep1b bound with both acidic and basic groups, and in five trajectories (1, 4, 6, 7, and 8), this is also observed frequently for class 1 during the whole simulation, indicating that Pep1b can bind to helical Ab with both acidic and basic functional groups at the same time from the beginning to the end of the simulation. In three of these trajectories (1, 4, and 7), Ab maintained its helical conformation and had not unfolded by the end of the simulation ( Table 1, group A).  As mentioned above, the group A trajectories exhibit nonunfolding or refolding of Ab. In one of the group A trajectories of each complex, trajectory 1 of the Ab-Dec-DETA complex and trajectory 2 of the Ab-Pep1b complex, Ab refolded to a helical conformation after being highly unwound during part of the simulations (Fig. 7A and 7B). Dec-DETA was bound to Ab during the first partial unfolding (8-11 ns) and refolding (11-12 ns) events, and during the first half of the second partial unfolding event (13-16 ns) but not during the second refolding event (16-17 ns) in trajectory Ab-Dec-DETA-1 (Fig. 7A). Pep1b was bound to Ab during the partial unfolding event (11-15 ns) except for a short break (12.5-13.5 ns), and was bound to Ab during the refolding event (15-16 ns) in trajectory Ab-Pep1b-2 (Fig. 7B). By visual inspection, we found that the charged functional groups of both Dec-DETA and Pep1b formed constant polar contacts with the charged sidechains of Ab when the ligands were bound to Ab during the partial unfolding and refolding periods, whereas the nonpolar contacts were intermittent.
According to our previous study [17], the Ab central helix does not completely unfold in cases where any of the three steps of the three-step mechanism, which was proposed for the complete unfolding of the Ab central helix, is missing: 1) sufficient loss of ahelical backbone hydrogen bonds, 2) strong interactions between nonpolar sidechains, and 3) strong interactions between polar sidechains. Here we observed that Ab did not completely unfold due to the lack of steps 3 and 2 in the first and second partial unfolding events, respectively, in trajectory Ab-Dec-DETA-1, and due to the lack of step 3 in the partial unfolding event in trajectory Ab-Pep1b-2.
These data suggest that strong inter-molecular interactions between the ligand polar groups and the Ab polar sidechains prevent intra-molecular interactions between the Ab polar sidechains, thus blocking the third step of the unfolding mechanism in trajectories Ab-Dec-DETA-1 and Ab-Pep1b-2. In this way Ab is inhibited from complete unfolding and instead Ab refolding is facilitated.

Ligand-Binding to Unwound Ab
As shown above, both ligands were able to bind to the unwound Ab structures in the peptide-conformation class 3, and long durations of the ligand-binding for class 3 were more frequent for the Ab-Dec-DETA complex than for the Ab-Pep1b complex (Fig. 7). This result suggests that both ligands, particularly Dec-DETA, have the possibility of being involved in the polymerization which occurs after the unfolding of the Ab central helix. To examine how the ligands interact with unwound Ab, we analyzed the ligand-binding events for class 3 in each individual trajectory in detail. Details of two Ab-Dec-DETA trajectories and one Ab-Pep1b trajectory, which exhibit long durations of the ligandbinding for class 3, are described below. Note that similar features were observed in the other trajectories of each Ab-ligand simulation.
In trajectory 3 of the Ab-Dec-DETA simulation, R g of Ab reaches a peak (R g $7.5 Å ) at around 17 ns (Fig. 8A), and one or two HBs between Dec-DETA and Ab are formed at the time (Fig. 8B). At around the time of the R g peak, b-strand-like forms of Ab bound by Dec-DETA were observed, and a typical structure of these forms was obtained at 16.96 ns (Fig. 8C). In this structure, two HBs are formed between Ab and Dec-DETA (Table 4), and the hydrocarbon sidechains of Ab are located close to the hydrocarbon chain of Dec-DETA (The Cc1(V18)-C5(Dec-DETA), Cc1(V18)-C6(Dec-DETA), and Cc2(V18)-C8(Dec-DETA) distances are 3.97, 3.93, and 4.00 Å , respectively.).
In trajectory 4 of the Ab-Dec-DETA simulation, R g of Ab reaches a peak at around 14 ns (Fig. 9A), and at least two HBs between Dec-DETA and Ab are formed at the time (Fig. 9B). bstrand-like forms of Ab bound by Dec-DETA were observed at around the time of the R g peak, and a typical structure of these forms was obtained at 14.29 ns (Fig. 9C). In this structure, four HBs are formed between Ab and Dec-DETA (Table 4) In both Ab-Dec-DETA structures obtained at the times of the R g peaks in trajectories 3 and 4, the hydrocarbon chain of Dec-DETA is located along the backbone of b-strand-like Ab, and thus, b-strand-like Ab and Dec-DETA form parallel conformations ( Fig. 8C and 9C). The parallel conformations of the Ab-Dec-DETA complex can be formed, due to the non-bulky conformation of Dec-DETA.
In trajectory 5 of the Ab-Pep1b simulation, R g of Ab reaches peaks at around 5 and 9 ns (Fig. 10A). The number of HBs between Pep1b and Ab is more than four at around 5 ns and is less than four at around 9 ns (Fig. 10B). These data show that Pep1b is tightly bound to the highly unwound or elongated Ab at around the time of the first R g peak but not at around the time of the second R g peak. After the first and second R g peaks, decreases in R g are observed together with decreases in RMSD (Fig. 10A), and several HBs between Pep1b and Ab are formed at these times (Fig. 10B), showing that Pep1b is bound to Ab which adopts compact forms at these times.
b-strand-like forms of Ab bound by Pep1b were not observed at around the times of both R g peaks, and instead, bent forms of Ab bound by Pep1b were observed. Typical structures of these forms observed at around the times of the first and second R g peaks were obtained at 5.15 and 9.12 ns, respectively (Fig. 10C). After the times of the first and second R g peaks, compact and partially helical forms of Ab bound by Pep1b were observed, and typical structures of these forms were obtained at 6.36 and 10.47 ns (Fig. 10C).
At the time of the first R g peak (5.15 ns), seven HBs are formed between Ab and Pep1b (Table 4), and the H14 imidazole ring of Ab is located close to the indole ring of Pep1b (The Cd2(H14)-C15(Pep1b) and Cd2(H14)-C20(Pep1b) distances are 3.27 and 3.42 Å , respectively.). The backbone of Ab is bent by the electrostatic interactions and by the auxiliary van der Waals interactions (Fig. 10C). After the time of the first R g peak (6.36 ns), six HBs are formed between Ab and Pep1b (Table 4), and the K16 sidechain and the F20 benzene ring of Ab are located close to the indole ring of Pep1b (The Cc(K16)-C19(Pep1b) and Cc(F20)-C19(Pep1b) distances are 3.69 and 3.93 Å , respectively.). The helical form of the backbone of Ab is partially (Q15-A21) reconstructed by the electrostatic interactions and by the auxiliary van der Waals interactions (Fig. 10C). At the time of the second R g peak (9.12 ns), two HBs are formed between Ab and Pep1b (Table 4), and the H13 imidazole ring and the F20 benzene ring of Ab are located close to the indole ring of Pep1b (The Cd2(H13)-C16(Pep1b) and Cc(F20)-C20(Pep1b) distances are 3.63 and 3.66 Å , respectively.). The backbone of Ab is partially (H13-F20) bent by the van der Waals interactions, though the backbone of Ab is partially (A21-S26) elongated (Fig. 10C). After the time of the second R g peak (10.47 ns), four HBs are formed between Ab and Pep1b (Table 4), and the H13 imidazole ring and the F20 benzene ring of Ab are located close to the indole ring of Pep1b (The Cc(H13)-C17(Pep1b) and Ce1(F20)-C20(Pep1b) distances are 3.82 and 3.34 Å , respectively.). The helical form of the backbone of Ab is partially (Q15-A21) reconstructed by the electrostatic interactions and by the auxiliary van der Waals interactions (Fig. 10C).
As shown in the Ab-Pep1b structures obtained in trajectory 5, the basic and acidic functional groups of Pep1b can simultaneously interact with the sidechains of the acidic and basic residues of Ab, respectively. In addition, the aromatic ring of Pep1b can at the same time interact with the aromatic rings of Ab. Ab therefore cannot easily convert to a b-strand-like form because of these electrostatic and van der Waals interactions. Even if Ab would be assumed to be a b-strand-like form, parallel conformations of the Ab-Pep1b complex cannot be formed, due to the bulky conformation of Pep1b.

Discussion
The effects of the two ligands (Dec-DETA and Pep1b) on the stability of the Ab central helix (residues 15-24) were investigated by using MD simulations. Detailed information on structural changes upon loss of helicity in the presence of the ligands was also examined, which might explain the observed difference in structures of Ab fibrils in the presence of Dec-DETA or Pep1b.
As indicated mainly by the Ab backbone RMSD vs the initial structure and by the existence of aHBs of Ab, the Ab central helix completely unfolded by the end of the simulation in three out of ten trajectories in the absence of a ligand, whereas it completely unfolded in only one out of ten trajectories in the presence of Dec-DETA and did not completely unfold in any of ten trajectories in the presence of Pep1b. Compared to Ab alone, the probability of the Ab helical state (more than 2/3 of all the aHBs are formed) during the second half of the simulations is 1.3 and 1.5 times higher for Ab in the presence of Dec-DETA and Pep1b, respectively. It was thus indicated that the stability of the Ab central helix was increased by both ligands, in agreement with the experimental data [12]. It was also indicated that the ability of Pep1b to stabilize the Ab central helix is higher than that of Dec-DETA, which was not shown in the previous experimental study [12]. Figure 10. Structural changes of trajectory 5 of the Ab-Pep1b system. The RMSD and R g of the Ab middle region (A) and the number of HBs between Ab and Pep1b (B) are shown. The structures obtained at 5.15 ns (with large RMSD (4.10 Å ), large R g (7.56 Å ), and seven HBs), at 6.36 ns (with medium RMSD (3.64 Å ), small R g (6.16 Å ), and six HBs), at 9.12 ns (with large RMSD (4.36 Å ), large R g (7.97 Å ), and two HBs), and at 10.47 ns (with medium RMSD (3.45 Å ), small R g (6.56 Å ), and four HBs) are also shown (C). doi:10.1371/journal.pone.0030510.g010 The analysis of the ligand-binding events clearly showed that Pep1b binds to the Ab central helix longer time than Dec-DETA does. A main reason for this is that Pep1b has both basic and acidic functional groups which can simultaneously bind to the acidic and basic residues of Ab, respectively, whereas Dec-DETA has only the basic functional groups. The inter-molecular interactions between the Ab polar residues and the ligand polar functional groups are important in stabilizing the Ab central helix, because they can prevent intra-molecular interactions between the Ab polar residues that induce complete unfolding of the Ab central helix [17]. An additional reason would be that Pep1b includes a centrally placed aromatic ring which straddles the Ab middle nonpolar part (residues 17-21) when the basic and acidic functional groups of Pep1b simultaneously bind to the acidic and basic residues of Ab, respectively. The inter-molecular interactions between the Ab middle nonpolar part and the ligand nonpolar part are likely to be important in stabilizing the Ab central helix, since the Ab middle nonpolar part includes the three nonpolar residues (VFF) that have low a-helical propensities and high b-strand propensities [39,40].
This analysis also showed that both ligands can bind to highly unwound or elongated forms of Ab. Dec-DETA was found to be able to form parallel conformations with b-strand-like forms of Ab. In contrast, Pep1b was found not to be able to form parallel conformations with b-strand-like Ab, due to the bulky conformation of Pep1b, and instead, Pep1b was found to bend unwound Ab by the charge-charge interactions and by interactions between the aromatic rings. Therefore, it may be suggested that Dec-DETA could be included upon formation and extension of b-sheets to Ab fibrils while being sandwiched between the two b-strands (residues 18-26 and 31-42) or being associated with the surface of a b-sheet, thus giving rise to fibrils with an alternative structure. On the other hand, Pep1b bound to unwound Ab may disturb the extension of b-sheets.
To summarize, it appears that Pep1b is somewhat more effective in stabilizing the Ab central helix than Dec-DETA. In addition, the difference in conformations between the unwound-Ab complexes bound by Dec-DETA and by Pep1b could be a reason why Ab incubated with Dec-DETA and with Pep1b form thicker-than-normal and shorter-than-normal fibrils, respectively, as reported by the previous experimental study [12], though the physical and physiological consequence of Dec-DETA containing alternative fibrils in vitro and in vivo is unknown. Hence, our study indicates that, compared to Dec-DETA-like ligands, Pep1b-like ligands, which are capable of having charge-charge interactions with both the acidic and basic residues of the Ab middle region, additional hydrophobic interactions with the Ab middle nonpolar part, and bulky conformations, appear to be more effective in inhibiting unwinding of helical Ab and also in preventing subsequent association of unwound Ab. Figure S1 Contact map of the Ab-Dec-DETA complex at 310 K. The probability (0.0#P,1.0) of the contact between the center of geometry of sidechain heavy atoms of each Ab residue and each Dec-DETA heavy atom is colored (white to blue grids). The probability was calculated using the data obtained from the whole simulation of one trajectory. The Ab residues and Dec-DETA atoms corresponding to the X and Y-axis numbers, respectively, are listed below the map. (TIFF) Figure S2 Contact map of the Ab-Pep1b complex at 310 K. The probability (0.0#P,1.0) of the contact between the center of geometry of sidechain heavy atoms of each Ab residue and each Pep1b heavy atom is colored (white to blue grids). The probability was calculated using the data obtained from the whole simulation of one trajectory. The Ab residues and Pep1b atoms corresponding to the X and Y-axis numbers, respectively, are listed below the map. (TIFF)