SEARCH

SEARCH BY CITATION

Keywords:

  • Alzheimer's disease;
  • amyloid beta peptide (Aβ);
  • oligomerization;
  • copper;
  • molecular dynamics simulations

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Conclusions
  7. Methods
  8. References

The formation of fibrils and oligomers of amyloid beta (Aβ) with 42 amino acid residues (Aβ1–42) is the most important pathophysiological event associated with Alzheimer's disease (AD). The formation of Aβ fibrils and oligomers requires a conformational change from an α-helix to a β-sheet conformation, which is encouraged by the formation of a salt bridge between Asp 23 or Glu 22 and Lys 28. Recently, Cu2+ and various drugs used for AD treatment, such as galanthamine (Reminyl®), have been reported to inhibit the formation of Aβ fibrils. However, the mechanism of this inhibition remains unclear. Therefore, the aim of this work was to explore how Cu2+ and galanthamine prevent the formation of Aβ1–42 fibrils using molecular dynamics (MD) simulations (20 ns) and in vitro studies using fluorescence and circular dichroism (CD) spectroscopies. The MD simulations revealed that Aβ1–42 acquires a characteristic U-shape before the α-helix to β-sheet conformational change. The formation of a salt bridge between Asp 23 and Lys 28 was also observed beginning at 5 ns. However, the MD simulations of Aβ1−42 in the presence of Cu2+ or galanthamine demonstrated that both ligands prevent the formation of the salt bridge by either binding to Glu 22 and Asp 23 (Cu2+) or to Lys 28 (galanthamine), which prevents Aβ1−42 from adopting the U-characteristic conformation that allows the amino acids to transition to a β-sheet conformation. The docking results revealed that the conformation obtained by the MD simulation of a monomer from the 1Z0Q structure can form similar interactions to those obtained from the 2BGE structure in the oligomers. The in vitro studies demonstrated that Aβ remains in an unfolded conformation when Cu2+ and galanthamine are used. Then, ligands that bind Asp 23 or Glu 22 and Lys 28 could therefore be used to prevent β turn formation and, consequently, the formation of Aβ fibrils.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Conclusions
  7. Methods
  8. References

The accumulation of insoluble amyloid “plaques” and oligomers of beta amyloid peptide (Aβ) in the brain plays an important role in Alzheimer's disease (AD), which is one of the principal causes of dementia worldwide,[1] causing oxidative stress and inflammation in specific areas of the brain.[2] Amyloid fibrils and oligomers are formed by a process called amyloidosis, in which the beta-amyloid peptide (Aβ) produces insoluble aggregates. Aβ is produced from the proteolytic cleavage of amyloid precursor protein (APP) by β and γ secretases or by α and γ-secretases to initiate amyloidogenic or nonamyloidogenic pathways, respectively.[3] The length of the Aβ peptide released by β-secretase varies from 39 to 43 amino acid residues, with the 42 amino acid residue peptide (Aβ1−42) being the most commonly found in fibrils. Aβ1−42 is composed primarily of hydrophobic amino acid residues at the C-terminal end,[4] which establish hydrophobic interactions during the oligomerization and fibrillization process. Furthermore, during these processes, the β-sheet conformation is defined and stabilized by the formation of an intrachain and interchain salt bridge between Asp 23/Glu 22 and Lys 28, increasing the ability of Aβ to nucleate and form fibrils.[5]

Because of the toxicity of Aβ1–42 oligomers and fibrils, the prevention of amyloidosis could have an enormous impact on the treatment of AD. Several studies have explored the use of metals and small compounds as oligomerization inhibitors, but a molecular-level understanding of the inhibition mechanism remains unclear. One of the metals that has attracted attention is copper (Cu2+), whose concentration is threefold greater in AD patients than in normal patients, with concentrations between 340 and 400 µM.[6, 7] The interaction between Aβ-fibrils and Cu2+ is important not only because of its role in amyloidosis but also because it has been associated with the generation of oxidative stress.[8] Therefore, several experimental and computational methods have focused on the analysis of the interaction between Aβ1−42 and Cu2+. However, although some research has shown that Aβ contains amino acid residues that can interact with Cu2+,[9] the effects of these interactions with respect to the Aβ1−42 oligomerization process have been contradictory. Some reports have demonstrated that the presence of Cu2+ favors fibril formation,[10] while other recent studies suggested that Cu2+ might inhibit Aβ1−42 fibril formation.[7] In addition, numerous compounds have been found to reduce Aβ1−42 aggregation in vitro.[11, 12] However, the usefulness of these compounds as inhibitors of Aβ1−42 aggregation remains speculative due to their lack of specificity and/or unknown mechanism of action. For example, galanthamine (Reminil®; AChE inhibitor) has been recently demonstrated to interact with soluble Aβ to inhibit the formation of the toxic oligomeric species in a concentration-dependent manner, but its mechanism of interaction has not been well established.[13]

Therefore, the elucidation of the mechanisms of Cu2+ and galanthamine inhibition of Aβ1−42 oligomerization may be useful in the rational design of drugs for AD patients. This study therefore aimed to analyze the Cu2+ and galanthamine interactions with Aβ1−42 using theoretical and experimental studies. For this purpose, we performed in silico studies; first, an Aβ1−42 monomer in the α-helix conformation with and without Cu2+ or galanthamine was submitted to molecular dynamics (MD) simulations to evaluate the conformational change. Subsequently, docking studies were performed using the conformer of Aβ1−42 obtained at 20 ns from the MD simulations to evaluate the interactions between the Aβ1−42 monomers and compare them with those of Aβ17−42, which forms the Aβ oligomers. The inhibition of the Aβ1−42 fibril formation by Cu2+ and galanthamine was assessed in vitro using ThT fluorescence, whereas the predominant secondary conformation was investigated using circular dichroism (CD) measurements.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Conclusions
  7. Methods
  8. References

142 sequence selection

Eleven Aβ sequences were identified from the Protein Data Bank (PDB) [Fig. 1(A)]. These sequences were submitted to multiple alignment. We observed that the sequences did not contain mutations; however, some sequences were incomplete, having only the residues 1–16 (PDB code: 2BP4) or lacking of these residues (PDB codes: 1QYT, 1QXC, and 1QWP), which belong to the disordered region (DR), because the Aβ1–42 structure has been divided into two major regions [Fig. 1(B)]. The first region comprises residues 1–17 and is characterized by its high degree of flexibility. The second region comprises residues 18–42 and is commonly known as the β-folded region (βR); it contains two intermolecular, parallel, in-register β-sheets formed from residues 18–26 (β1) and 31–42 (β2), which are connected by a turn formed by residues 27–30[14, 15] [Fig. 1(B)]. As depicted in Figure 1(A), several sequences, such as 1QYT, 1QXC, and 1QWP, contained amino acids involved in the turn and the β2 region. However, other structures, such as 1Z0Q, 2BEG, IIY3, and 1AML, possessed complete sequences (Abeta1-42). We chose to study 1Z0Q and 2BEG because these structures have been most frequently quoted in the literature.

image

Figure 1. The Aβ1−42 amino acid sequence and structure. (A) The alignment of Aβ sequences obtained from the Protein Data Bank. (B) A schematic representation of Aβ1−42. Residues 1–17 comprise the disordered region (DR), residues 18–26 comprise the β1 region, residues 27–30 comprise the turn region, and residues 31–42 comprise the β2 region. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

Molecular dynamics simulations

MD simulation of Aβ142 in the α-helix conformation

The Aβ1−42 structure selected for the MD simulation was that with PDB code 1Z0Q, which has an α-helix conformation [Fig. 2(A)]. It gradually acquired a characteristic U-shape after 5 ns [Fig. 2(B)], forming and stabilizing the spin portion (amino acid residues 27–30), as depicted in Figure 2(C,D). The average values of the root-mean-square deviation (RMSD), the radius of gyration (Rg) and the root-mean-square fluctuations (RMSF) of the Aβ1−42 simulation were calculated, taking into account the entire MD simulation. As displayed in Figure 2(E), the maximum RMSD value for all of the Cα atoms was ≅9 Å, and peptide convergence was reached after 6 ns, which allowed a stationary state in which the RMSD approached its maximum. The Rg values, which are a measure of the compactness of the protein, were highest during the first 6 ns and achieved convergence [Fig. 2(F)]; these results are in agreement with the views in Figure 2(B–D), where Aβ1−42 acquires the characteristic U-shaped architecture and the β-turn after 5 ns. These structural changes related to the secondary structure can be analyzed using the RMSF, which exhibit peaks that correspond to the average atomic mobility of the backbone. Based on the RMSF results, a high degree of variability was observed in the residues that comprise the turn and the β2 region, which implies that a conformational change occurred in the Aβ1−42 structure, as shown in Figure 2(G). The formation of a salt bridge between Asp 23 and Lys 28 enabled the conformational change of Aβ1–42, producing the β-turn. As displayed in Figure 2, the distance between these amino acids diminished considerably, from 22.02 Å at 0 ns [Fig. 2(A)] to 17.49 Å at 5 ns [Fig. 2(B)], 13.44 Å at 15 ns [Fig. 2(C)], and 5.6 Å at 20 ns [Fig. 2(D)]. The salt bridge formation allowed the amino acids at the turn to gradually acquire a β-sheet conformation. As observed in Figure 3(A), these amino acids were in an α-helix conformation at the beginning of the simulation, and Gly 29 and Lys 28 achieve a β-sheet conformation after 20 ns of simulation [Fig. 3(B)].

image

Figure 2. Results obtained from the Aβ1−42 MD simulation. The top views of the conformers obtained at different MD simulations times: (A) 0 ns, (B) 5 ns, (C) 15 ns, (D) 20 ns, (E) RMSD, (F) Rg, and (G) RMSF. Notably, at 5 ns, Aβ1−42 acquires a characteristic β-turn that remains constant throughout the simulation, and near this time point the RMSD reaches convergence. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

image

Figure 3. Ramachandran plots of Aβ1–42 conformers obtained from the MD simulations. Aβ1–42 in an α-helix conformation at time 0 ns (A) and 20 ns (B); Aβ1–42 in the presence of Cu2+ (C) and galanthamine (D) after 20 ns of simulation. The amino acids that comprise the turn have been identified. As shown in the figure, the amino acids gradually adopt a β-sheet conformation, but in the presence of Cu2+ or galanthamine, the conformational change does not occur. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

MD simulation of Aβ142 in an α-helix with Cu2+

Figure 4 illustrates the top view of the results from the MD simulation for Aβ1–42 in the presence of Cu2+. It is possible to observe that the Aβ1–42 remained in the α-helix conformation for the entire duration of the MD simulation [Fig. 4(A–D)] and that the U shape was not observed. Furthermore, because the Cu2+ and His residues did not interact, the results from this MD simulation contradict those reported in the literature.[9, 16, 17] Consequently, we evaluated the amino acid residues that interacted with Cu2+, Asp 23, and Glu 22, as displayed in Figure 4(B–D). The distance between Cu2+ and Asp 23 was lower (2.01 Å) compared with that between Cu2+ and Glu22 (4.81 Å) at the end of the simulation [Fig. 4(D)].

image

Figure 4. Results obtained from the Aβ1−42 α-helix MD simulation in the presence of Cu2+. The top views of the conformers at various simulation times are shown. (A) Time 0, (B) 5 ns, (C) 15 ns, (D) 20 ns, (E) RMSD, (F) Rg, (G) RMSF. As evident from the results, no conformational changes were observed, in contrast to Aβ1−42 without Cu2+. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

The coordination geometry of Cu2+ is typically tetrahedral, where Cu2+ can interact with up to four water molecules; however, it has also been reported that Cu2+ can interact with amino acid residues other than His.[16, 17] Thus, for Aβ1−42 in the α-helix conformation, Asp 23 did not interact with Lys 28, which allowed the negatively charged zone (Asp23 and Glu22) to interact with Cu2+. Furthermore, the carboxylate group can act as a bidentate ligand in coordination with copper, in which the tetrahedral geometry is conserved.

The RMSD, Rg, and RMSF values of the Aβ1–42–Cu2+ interaction during the MD simulations were calculated considering the entire dynamic simulation. Figure 4(E) displays the RMSD values for all of the Cα atoms. The maximum RMSD was 5 Å; however, a variation of 2 Å was observed, indicating that conformational changes in the model occurred during the MD simulation. After 10 ns, the RMSD value approached the maximum and reached convergence, as shown in Figure 4(E). The Rg values were without slight variations throughout the simulation [in Fig. 4(F)]. Based on the RMSF results, a high degree of variability was observed for the entire Aβ1−42 structure, as shown in Figure 4(G). The Asn 27, Asp 23, and N-terminal residues exhibited the highest mobility. No great conformational changes were observed in this MD simulation [Fig. 4(A–D)], and the turn was not formed. This result corresponds with Figure 3(C), where after 20 ns of simulation in the presence of Cu2+, the amino acids at the turn location remain in an α-helix conformation.

MD simulation of Aβ142 in a β-folded conformation with Cu2+

To evaluate the effect exerted by Cu2+ on Aβ1–42 in the β-folded conformation, a 20-ns MD simulation was performed. The coordinates for Aβ1–42 were taken from the results of the Aβ1–42 simulation without Cu2+ at 13 ns (structure refinement). Although the Cu2+ was located near the reported specific amino acid interaction (His 6, 13, and 14), it exhibited a tendency to move toward the negatively charged side chains of Glu 22 and Asp 23 [Fig. 5(A–D)]. The interaction between Cu2+ and Glu22 and Asp23 prevented the secondary conformational change as well as the gradual loss of the U shape [Fig. 5(D)].

image

Figure 5. Results obtained from the MD simulation of Aβ1−42 in the β-sheet conformation with Cu2+. The top views of the conformers at various times are shown. (A) Time 0, (B) 5 ns, (C) 15 ns, (D) 20 ns, (E) RMSD, (F) Rg, (G) RMSF. As evident from the figure, the presence of Cu2+ induces the loss of the β-sheet conformer of Aβ1−42 and the formation of the α-helix conformer. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

Figure 5(E) demonstrates that the maximum RMSD value for all of the Cα atoms was 8 Å, with the occurrence of wide variations that demonstrate that the conformational changes were induced by Cu2+. In contrast, the Rg value was not stable at any point during the simulation [Fig. 5(F)]. The RMSF results indicate a high degree of variability, especially for the DR [Fig. 5(G)].

MD simulation of Aβ1–42 with galanthamine

The top view in Figure 6 displays the results from the MD simulation of Aβ1–42 with galanthamine. Aβ1–42 was in the α-helix conformation at the beginning of the simulation [Fig. 6(A)] and gradually acquired a U-shaped structure [Fig. 6(B,C)]. As shown in Figure 6(D), the principal interaction between galanthamine and Aβ1–42 involved the π-cation with Lys 28 whereas the N atom positively charge of galanthamine make a salt bridge with Asp 23.

image

Figure 6. Results obtained from the Aβ1−42-galanthamineMD simulation. The top views of the conformers at various MD simulation times: (A) time 0, (B) 5 ns, (C) 15 ns, (D) 20 ns, (E) RMSD, (F) Rg, (G) RMSF. Galanthamine did not interact with Aβ1−42 during the simulation. Significantly, at 20 ns, galanthamine was observed to form a π-cation interaction with Lys 28. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

The RMSD, Rg, and RMSF values of the interaction between Aβ1–42 and galanthamine were calculated for the entire MD simulation. As displayed in Figure 6(E), the maximum RMSD value for all of the Cα atoms was 9 Å. The protein reached convergence at ≅6 ns, where the RMSD approached its maximum, which was then maintained throughout the remainder of the MD simulation. The Rg was stable at 12 Å in the complete simulation [Fig. 6(F)]. Based on the RMSF results, the residues that comprised the DR exhibited a large variability [Fig. 6(G)]. The Ramachandran plot at 20 ns revealed that the amino acids located at the turn remained in an α-helix conformation [Fig. 3(D)].

Docking studies

For docking studies, we used the 2BEG structure (residues 17–42) and a snapshot conformer of Aβ1–42 obtained by MD simulations at 20 ns (PDB code: 1Z0Q, a complete 3D structure). Both structures have the minimal sequence able to form the oligomers.[18] Therefore, the protein–protein docking was first performed with two monomers of each structure to build a dimer (Fig. 7, Set 1, in blue). From these studies, according to the Clus Pro 2.0 server, four main interactions were observed between the monomers, which were induced by (a) all types of interactions, known as balanced; (b) hydrophobic forces; (c) electrostatic forces; and (d) Van der Waals interactions (VdW).[19, 20]

image

Figure 7. 1−42 oligomers obtained by docking using using a snapshot of Abeta1-42 obtained by MD simulations at 20 ns (PDB code: 1Z0Q). In each set of figures, the letter represents a type of interaction: (A) all types of interactions, known as balanced; (B) hydrophobic interactions; (C) electrostatic forces; (D) van der Waals forces. For the dimer (Set 1, in blue), trimer (Set 2, in green), tetramer (Set 3, in pink), and pentamer (Set 3, in orange). (E) A schematic representation of the formation of the salt bridge between Asp 23 and Lys 28 in the dimer. (F) A schematic representation of the formation of another salt bridge between Asp 7 and Lys 16 in the pentamer. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

An analysis of the interactions between the oligomers obtained using the conformer from the MD simulation of the 1Z0Q and 2BGE structures revealed different orientations, including dimers (Fig. 7, Set 1 in color blue), trimers (Fig. 7, Set 2 in color green), tetramers (Fig. 7, Set 3 in color pink), and pentamers (Fig. 7, Set 4 in color orange), that depended on the type of interaction present. However, only the oligomers formed by hydrophobic interactions exhibited unidirectional growth along the axis.

These oligomers were subsequently used to analyze interchain interactions, such as hydrogen bonds, that formed between the backbone of residues Val 39–Val 18, Asp 23–Leu 34, Lys 28–Val 36, Glu 22–Met 35, and Phe 20–Gly 37 and that were constant in all of the oligomers (Table 1). Importantly, other interactions were observed in the oligomer (Gly 38–Phe19, Gly 38–Val 39, Ile 32–Gly25, Leu 17–Val 40 for the trimer, and Val 36–Ala 21 for the tetramer). Furthermore, salt bridge (Asp23–Lys28) formation was observed in both structures (2BEG and 1Z0Q conformer 20 ns from MD), and the formation of these salt bridges induced interactions between neighboring hydrophobic side chains [Fig. 7(E)]. As outlined in Table 1 (1Z0Q conformer), this salt bridge was present in all of the oligomers and was shorter in length when more Aβ chains were present. However, in the structure 1Z0Q, additional interactions between the monomers, such as electrostatic interactions between Asp 7 and Lys 16, were observed due to the presence of the DR [Fig. 7(F)]. In addition, in the 1Z0Q structure, the interchain hydrogen bonds between the backbone of the amino acid residues Met35–Val40, Arg5–Tyr10, Glu11–Hys6, Val36–Ile41, Ile31–Val36, and Phe20–Gln15 remained constant in all of the oligomers; importantly, other hydrogen bonds were added and remained constant in the subsequent higher order oligomer (Ala30–Met35 in the trimer and Phe19–His14 and Phe19–His14 in the tetramer).

Table 1. Interactions Encountered in the Aβ1–42 (PDB code 1Z0Q) Oligomers Obtained by Docking Studies
DimerTrimerTetramerPentamer
  1. a

    Distances are presented in Å.

Salt BridgeSalt BridgeSalt BridgeSalt Bridge
Asp 23-Lys 28 1.88aAsp 1-Lys 28 2.77aAsp 23-Lys 28 2.98aAsp 23-Lys 28 1.79a
Asp 23-Lys 28 2.44a Asp 23-Lys 28 2.43aAsp 23-Lys 28 1.28a
Lys 16-Asp 7 1.77aHYDROGEN BONDSAsp 23-Lys 28 3.28aAsp 23-Lys 28 2.84a
 Ile41 H-Val36 O (1.81)Asp 23-Lys 28 2.98aAsp 7-Lys 16 1.7a
HYDROGEN BONDSIle 31 O-Val36 H (1.81)Asp7-Lys16 1.82aAsp 7-Lys 28 3.13a
Met 35 O- Val 40 H (1.86)Val40 H-Met35 O (1.89)Lys16-Asp7 2.54aAsp 7-Lys 16 1.7a
Glu 11 O- Tyr 10 (1.99)Arg5 H-Asp10 O (2.0) Asp 7-Lys 16 1.28a
Asn27 H–Val 24 O (2.02)Asp7 H-Glu3 O (1.95)HYDROGEN BONDSAsp 7-Lys 16 2.84a
Ser26 H-Glu22 O (1.94)Lys28 H-Val24 O (1.97)Ile41 H-Val38 O (1.81) 
Gln15 O-Tyr10 H (1.92)Val18 O-Asp23 H (1.88)Val36 H-Ile31 O (1.81)HYDROGEN BONDS
Tyr10 O-Arg5 H (1.87)Met35 H-Ala 30 O (1.87)Val40 H-Met35 O (1.89)Val24 O-Asn27 H (2.01)
Asp7 H-Glu3 O (1.73)Val 40 H-Met35 O (1.91)Arg5 0-Tyr10 H (1.86)Ser26 H-Glu22 O (1.95)
Glu11 H-His6 O (1.95)Lys16 O-Ala21 H (1.86)Glu22 O-Asp7 H (1.73)Phe19 O-Val24 H (1.86)
Tyr10 H-Glu11 O (1.91)Phe20 H-Gln15 O (1.91)Asp1 H-Glu3 O (1.95)Asp1 O-Arg5 H (1.92)
Lys16 H-Glu11 O (1.9)Val24 H-Phe19 O (1.86)Hys6 O-Glu11 H (1.96)Ile31 O-Val36 H (1.82)
Val36 O-Ile41 H (1.82)His6 O-Glu11 H (1.85)Glu11 O-Lys16 H (1.96)Val36 O-Ile41 1.81 (1.81)
Met35 O-Val40 H (1.89)His6 H-Tyr10 O (2.03)Tyr10 O-Gln15 H (1.92)Phe19 H-His14 O (1.84)
Ile31 O-Val36 H (1.81)Ser8 O-His13 H (1.93)Lys16 O-Ala21 H (1.86)Gln15 H-Tyr10 O (1.92)
Ala21 H-Lys16 O (1.86)Val40 H-Met35 O (1.83)Gln15 O-Phe20 H (1.94)Met35 O-Val40 H (1.89)
Phe20 H-Gln15 O (1.91)Glu11 O-Lys6 H (1.89)Tyr10 H-Glu11 O (2.03)Tyr10 H-Arg5 O (1.88)
Val24 H-Phe19 O (1.86)His6 H-Glu11 O (1.95)Phe19 H-Hys14 O (2.85)Met35 O-Val40 H (1.86)
 Glu3 O-Asp7 H (1.73)Val24 H-Phe19 O (1.86)Phe20 H-Gln15 O (1.91)
 Arg5 H-Asp1 O (1.88)Ser26 H-Glu22 O (1.96)Glu11 O-Tyr10 H (2.03)
 His14 H-Gly9 O (1.87)Asn27 H-Val24 O (2.01)Ala21 H-Lys16 O (1.96)
 Lys28 H-Val24 O (1.94)Val40 H-Met35 O (1.86)Val18 H-His3 O (1.88)
 Phe20 H-Gln15 O (1.85)Ile41 H-Val36 O (1.85)Phe19 H-His14 O (1.97)
 Glu22 O-His6 H (1.98)Phe4 O-Gly9 H (1.96)Lys16 H-Glu17 O (1.89)
  Asp7 O-Arg5 H (1.75)Arg5 H-Glu11 O (1.93)
  Gly37 H-Ile32 O (1.93)Glu11 H-His6 O (1.95)
  His13 O-Val18 H (1.97)Glu3 O-Asp1 H (2.01)
  Glu22 O-Lys28 H (1.75)His6 H-Ala2 O (1.85)
  Lys28 H-Glu22 O (1.81)Gly29 H-Ser26 O (2.01)
  Glu22 O-Ser26 H (1.9)Ser26 H-Glu22 O (1.87)
  Asp23 O-Lys28 H (1.65)Val40 H-Met35 O (1.92)
  Val24 O-Asn27 H (1.96)Met35 O-Val40 (1.9)
  Hys6 O-Glu11 H (1.93)Lys28 H-Asp7 O (1.81)
  Tyr10 O-Gln15 H (1.85)Glu3 O-Asp7 H (1.74)
  Met35 H-Ala30 O (1.95)His14 H-Gly9 O (1.92)
  Val18 O-Glu22 H (1.96)His6 H-Ala2 O (1.96)
  Lys16 H-Tyr10 O (1.63)Glu11 O-Tyr10 H (1.93)
  Lys16 H-Glu11 O (1.71)Tyr10 H-Glu11 O (1.91)
  Val40 H-Met35 O (1.86)Val12 H-Asp7 O (1.91)
  Leu34 H-Gly29 O (1.93)Val18 H-His13 O (1.9)
  Tyr10 H-Glu11 O (1.9)Gln15 O-His14 H (1.9)
  Val18 O-Asp23 H (1.94)Phe19 H-His4 O (1.95)
  His14 O-Phe19 H (1.94)Lys16 H-Glu11 O (1.88)
  Leu17 O-Glu22 H (1.92)Asn27 H-Val24 O (1.93)
  His13 O-Val18 H (1.98)Leu17 O-Ala21 H (1.92)
  Gly9 O-His14 H (1.97)Asp23 O-Asp14 H (1.94)
  Ser8 O-His13H (1.91)Ala30 O-Met35 H (1.94)
  Phe4 O-Gly9 H (1.94)Phe19 H-His14 O (1.82)
  His6 O-Glu11 H (1.93)Ser26 H-Glu22 O (1.92)
  Glu11 O-Lys16 H (1.94)Glu11 O-Lys16 H (1.72)

Evaluation of Aβ142 aggregation by thioflavin T fluorescence

Monitoring the in vitro formation of Aβ1–42 fibrils

The fluorescence signal of thioflavin T (ThT) has been well documented to increase when ThT is bound to amyloid structures; therefore, this signal can be used to evaluate the formation of misfolded protein aggregates.[21] For this experiment, nonfibrillated Aβ1–42 peptide was used as the starting material to monitor the fibrillization process. The experimental conditions for Aβ1–42 fibrillization have been described in previous reports.[7] The increase in the intensity of the fluorescence signal of ThT associated with Aβ1–42 fibrillization is depicted in Figure 8(A). The time course for the formation of the fibrils was also observed through continuous monitoring of the changes in the ThT fluorescence intensity at 480 nm over the course of 24 h; this time period is sufficient to allow fibril formation under incubation and stirring conditions. A rapid increase in the ThT fluorescence signal is demonstrated in Figure 8(B), which depicts a two-phase-growth curve that is typical of amyloid fibril formation. When Aβ1–42 was incubated without agitation, the fibrillization process was extremely slow, and the increase was only observed after prolonged incubation (data not shown).[22]

image

Figure 8. Effect of compounds on Aβ1−42 fibrillization, analyzed using the fluorescence intensity of ThT in the presence of Cu2+ and galanthamine after 24 h of incubation: (A) 0.9 µM1−42 (dotted line), 0.9 µmol/L Aβ1−42 in the presence of 20 µM Cu2+ (continuous line), 0.9 µM1−42 in the presence of 100 µmol/L galanthamine (dashed line). (B) Emission at 480 nm of Aβ1−42 (continuous line), Aβ1−42 and Cu2+ (dashed line), and Aβ1−42 and galanthamine (dotted line).

Download figure to PowerPoint

Effects of Cu2+ and galanthamine on the formation of Aβ142 fibrils

Both Cu2+ and galanthamine inhibited Aβ1–42 fibrillization, as depicted in Figure 8 (panels A and B). The addition of Cu2+ prevented Aβ1–42 fibrillization, as demonstrated by the small increase in the ThT fluorescence signal following Cu2+ addition. The addition of galanthamine also decreased the fibrillization process, but the effect was not as pronounced as that observed for Cu2+. These observations are in good agreement with the MD simulation results, in which Cu2+ exhibited a high affinity for the Glu 22 and Asp 23 side chains, as demonstrated by the proximity of Cu2+ to these amino acid residues throughout the duration of the simulation. In comparison, galanthamine interacted weakly with Lys 28 in Aβ1–42; however, this interaction was sufficient to prevent the fibril formation process, as demonstrated by the fluorescence results.

Circular Dichroism

Figure 9 displays representative CD spectra of Aβ1–42 alone and in the presence of Cu2+ or galanthamine following 1 h of incubation; the spectra were corrected for ellipticity using the CD spectra of Cu2+ or galanthamine. After the sample was incubated for 1 h under conditions that favored fibril formation, the CD spectrum revealed changes in the secondary structure, which indicated the formation of the βI conformation for Aβ1–42. When Aβ1–42 was incubated in the presence of either Cu2+ or galanthamine, the CD spectra in both cases indicated a very high proportion of βII content (68 and 67%, respectively) and a significant reduction in the proportion of βΙ (Table 2).

image

Figure 9. CD spectra of Aβ1–42 after 1 h of incubation alone (50 µM; continuous line) and in the presence of Cu2+ (20 µM; dotted line) or galanthamine (100 µM; dashed line).

Download figure to PowerPoint

Table 2. The Proportions of the Aβ42 Secondary Structure Induced by 20 µM Cu2+ and 100 µM Galanthamine, Sampled After 1 h of Incubation
 α (%)βΙ (%)βΙΙ (%)Error (%)
78763
Aβ−Cu2+2012685
Aβ−Gal1815672

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Conclusions
  7. Methods
  8. References

The formation of Aβ fibrils and oligomers is one of the major pathological hallmarks of AD.[23] These fibrils have been shown to produce reactive oxygen species (ROS) such as H2O2 in vitro as well as OH in the presence of transition metals such as Cu2+, establishing a permanent state of oxidative stress.[8] This oxidative stress affects the cholinergic transmission system, which explains the underlying cholinergic deficit in AD patients.[24]

Several efforts have been made to elucidate the mechanism of ROS production due to the interaction of Aβ fibrils and Cu2+;[8] however, the effect of Cu2+ and other metals on the oligomerization process is not well defined, and different authors have reported contradictory results. Some reports have demonstrated that Cu2+ favors the aggregation of Aβ1–42. However, in recent experiments, the incubation of Aβ1–42 in the presence of Cu2+ resulted in decreased Aβ fibril formation, although no explanation for these processes at the atomistic level has been provided.[11-13]

Although several theoretical and experimental studies have been conducted with the aim of explaining the interaction of Cu2+ and Aβ, several contradictory results have emerged.[16] These contradictions might be explained by the different experimental conditions used by different research groups, which have included different pH values,[23] different types of peptides, and differences in the form in which Cu2+ could be coordinated to Aβ;[17] all these conditions can influence Aβ aggregation.

MD simulations were performed for the 1Z0Q structure and Cu2+ to elucidate the process by which Cu2+ prevents Aβ1−42 fibril formation. From the results of this study, we demonstrated that, in the absence of Cu2+, Aβ suffered conformational changes. At 5 ns of the MD simulation, the Aβ adopted a U-turn shape that encouraged the interaction between Asp 23 and Lys 28, as reported previously.[25] This adoption of a U-turn shape also indicated the beginning of the transformation to a β-sheet structure, which is a crucial step in initiating the oligomerization process.

However, in the MD simulation of Aβ1–42 with Cu2+, we did not observe a significant change in conformation, and the Asp 23 and Lys 28 residues remained distant throughout the simulation. In fact, for nearly the entire simulation, Cu2+ interacted with Glu 22 and Asp 23. This behavior was likely observed because Cu2+ is a cation metal and is attracted to negatively charged molecules, such as the carboxylate groups present in the side chains of the amino acid residues in Aβ1–42. Importantly, some authors have reported that Cu2+ is associated with Aβ1–42 in the β-folded structure and is tetracoordinate with three amino acid residues, including His 6, His 13, and His 14, and a fourth component represented by a residual oxygen donor—possibly Asp or Tyr. However, these studies began with Aβ1−42 in a β-helix conformation, and the salt bridge between Asp 23 and Lys 28 was formed. In addition, these authors observed that the change in the β-sheet folding occurred within picoseconds and that the interaction with Cu2+ occurred following the conformational shift.[26] Furthermore, because Aβ1−42 is composed principally of hydrophobic amino acid residues, the amino acids that can interact with Cu2+ are limited to Glu 22 and Asp 23. Moreover, for Aβ1–42 in the β-sheet conformation, the presence of the salt bridge between Asp 23 and Lys 28 prevents the interaction of these residues, favoring the interaction of Cu2+ with the imidazole ring of histidine. In fact, the coordination of Cu2+ with the imidazole rings of histidine was demonstrated in Aβ1–42 fibrils, in which the β-sheet conformation is stabilized by the salt bridge between Asp 23 and Lys 28.[9] However, as reported in the MD simulation in this work, for Aβ1–42 (α-helix)–Cu2+, due to the absence of a salt bridge, the Cu2+ is free to interact with Glu 22 and Asp23. In addition, many of the studies in which Cu2+ was reported to interact with the histidine residues of Aβ were performed using Aβ1−16[27] Additionally, the RMSD plots and rotation of the Cα revealed the presence of significant instability in the Aβ1–42 chain, implying that this type of interaction promotes the destabilization of the peptide and prevents the formation of the salt bridge necessary for the transition to a β-sheet conformation. In fact, the Aβ1–42 (β-sheet)–Cu2+ simulation revealed that Cu2+ possesses the ability to destabilize the β-sheet conformation of Aβ1–42. This destabilization may be useful in explaining the results of the experiments in which the incubation of Aβ with Cu2+ inhibited Aβ aggregation.

Based on the MD simulation results for Aβ1–42 with galanthamine, galanthamine binds to Aβ1–42 in two different manners: the aromatic ring of galanthamine interacts with Lys 28 of Aβ1–42 via π-cation interactions, and the protonated tertiary amine interacts with Asp 23 via electrostatic interactions, preventing the formation of the salt bridge that was shown to form before the Aβ1–42 conformational change. Both types of interactions are capable of preventing the Aβ1–42 conformational change from α-helix to β-sheet; according to previous reports,[28] galanthamine interacts with the amino acids located at the turn portion of Aβ1–42, such as Lys 28. However, the affinity of Aβ1–42 for galanthamine is lower than that for Cu2+, as revealed by the higher concentration of galanthamine (100 µM) necessary to inhibit the oligomerization process[13] compared with Cu2+ (20 µM). This difference in affinity has been corroborated in this study, in which only Cu2+ was capable of decreasing the maximum ThT fluorescence following incubation with Aβ1–42.

In addition, for the docking studies, the structures of Aβ17–42 (PDB code 2BEG) and Aβ1–42 (conformer obtained from MD at 20 ns) were used to evaluate whether the structure obtained by MD could form oligomers in the same manner as the 2BGE structure and whether amino acid residues 1–16 were important in the salt bridge formation between Asp23 and Lys28. The results from the docking study between the monomers from the 1Z0Q conformer and the 2BEG structure revealed that the addition of consecutive Aβ monomers can occur in different ways, resulting in a wide variety of associations dependent on the main type of interactions present. Thus, the large number of interactions that Aβ1–42 can establish with other monomers with a small free energy allows the formation of different types of Aβ aggregates. Because these aggregates are strongly affected by their environment, the competing influences of the fragment interactions and the key nucleation sites may determine the polymorphic oligomer (and fibril) outcome, particularly with respect to the parallel/antiparallel backbone alignment.[29] However, not all types of aggregates are important, as demonstrated in the case of senile plaques, in which Aβ1–42 primarily forms fibrils in the toxic form.[23] Moreover, the process by which the oligomers form hydrophobic interactions most closely resembles the oligomerization process reported for the formation of Aβ fibrils; these types of interactions might explain the low hydrophilicity of Aβ following fibril formation. Notably, although the Aβ monomers from the 2BEG structure lacked the 1–16 region, the oligomers obtained were similar in structure, conformation and behavior to the oligomers obtained from the 1Z0Q structure, which might indicate that the DR is not essential in the oligomerization process. However, for the complete Aβ sequence (1Z0Q structure), more electrostatic interactions were observed in the presence of the DR.

Additionally, experimental CD studies confirmed the computational results, revealing that Cu2+ and galanthamine can prevent the Aβ1–42 conformational change and allow the predominance of amino acids in an unfolded conformation, as demonstrated by the resemblance of the spectra to the CD spectra of βII proteins.[30] This conformation prevents Aβ1–42 oligomerization, as demonstrated by the fluorescence results.

Based on these results, we concluded that three important characteristics should be considered in the rational design of Aβ1–42 fibrillogenesis inhibitors: the presence of hydrophobic groups, aromatic rings or charged groups that can interact with Glu 22 and Asp 23 or Lys 28—characteristics that galanthamine possesses because it contains a tertiary amine capable of being protonated at physiological pH; a positive charge capable of interacting via electrostatic interactions with Asp 23; and aromatic rings capable of interacting with Lys 28 via π-cation interactions. However, this molecule inhibits Aβ1–42 fibril formation only weakly due to its large size. Consequently, steric hindrance effects should be considered during the design of new molecules because these effects may contribute to the ineffectiveness of these molecules as inhibitors of Aβ1–42 fibril formation.

Conclusions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Conclusions
  7. Methods
  8. References

From in silico studies, we found that Cu2+ is able to prevent and revert the Aβ1−42 secondary-structure conformational change by interacting with Glu 22 and Asp 23 to prevent the formation of a salt bridge. In comparison, galanthamine, which bound to Aβ1−42 via π–cation interactions with Lys 28, also inhibited salt-bridge formation, but by a different mechanism. The inhibition of the Aβ1−42 fibrillation process was observed using ThT fluorescence. In addition, this inhibition was further confirmed using CD spectroscopy.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Conclusions
  7. Methods
  8. References

Aβ sequence selection

To use the most complete three-dimensional (3D) Aβ structure without mutations or missing residues, a search in the Protein Data Bank (PDB) (http://www.rcsb.org/pdb/home/home.do) was performed, which identified 11 structures (PDB codes: 2BP4, 2BEG, 1Z0Q, 1QYT, 1QXC, 1QWP, 1IYT, 1HZ3, 1AML, 1AMC, 1AMB). A multiple alignment was performed using the STRAP program (http://www.bioinformatics.org/strap/), and the results were edited to find similarities and differences in the sequences.

Molecular dynamics simulations

The native Aβ1−42 monomer was used for the MD simulations. The calculations were performed using the structure of Aβ1−42 (PDB code: 1Z0Q model 1) determined by Tomaselli et al.[31]

MD simulation of Aβ1−42 in an α-helix structure

MD simulations were performed with the NAMD 2.6 software package.[32] Hydrogen atoms were added with the psfgen command within the VMD program[33], and the structure was minimized using the steepest descent algorithm for 2000 steps using the CHARMM27 force field.[34] The resulting structure was immersed in water (10 Å TIP3 water model), and the charge was neutralized using 3 Na+ ions. The particle-mesh Ewald method[35] and periodic boundary conditions were applied to complete the electrostatic calculations. We used Nose-Hoover Langevin piston pressure control and maintained the temperature at 310 K.[36] The SHAKE[37] method was used to provide an integration time step of 2 fs while keeping all bonds to the hydrogen atoms rigid. The equilibration protocol involved 1500 minimization steps followed by 30 ps of MD at 0 K for the water and ions while freezing the entire protein. Once the minimization of the entire system was achieved, the temperature was increased from 10 to 310 K over 30 ps to ensure that it would continue to modify its volume with 30 ps of NTP dynamics.[36] As a final step, the NTV dynamics continued for 20 ns. The trajectory of the system was stored every 1 ps, and the simulations were analyzed by capturing several snapshots every 1 ns. The snapshots, RMSD, RMSF, and Rg were obtained using the carma program[38] based on the following equations. The RMSD was analyzed to determine whether the protein had undergone a conformational change because this value reflects the distance between pairs of the same atoms represented by δ with respect to time. However, the RMSF indicates the displacement of the particle, as observed in Eq. (2). Therefore, the RMSF average is taken over time to give a value for each particle, whereas the RMSD average is calculated over the particles to provide time-specific values.

  • display math(1)
  • display math(2)

All the computational work was performed using pmemd cuda on an Intel Core i7–980x 3.33 Ghz Linux workstation with 12 Gb of RAM, (2x) NVIDIA Geforce GTX530 video cards, and (1x) NVIDIA Geforce GTX580 video card.

MD simulations of Aβ1−42 -Cu2+

To perform MD simulations that included Cu2+, the parameters of Cu2+ were added to the topology (MASS 63.54600, RESI CU2 2.00) and parameter (CU 0.000000–0.250000 1.090000 ION) files. The coordinates of Cu2+ were taken from the crystallized structure of cytochrome C oxidase (PDB code: 1OCO),[39] and 1 Cu2+ ion was placed near the Hys 6, 13, and 14 amino acid residues because these residues have been reported to be capable of binding Cu2+. Hydrogen atoms were added with the psfgen command within the VMD program,[33] and the structures were minimized using the steepest descent algorithm for 2000 steps using the CHARMM27 force field. The structure was immersed in water (10 Å TIP3 water model), and the charge was neutralized using 1 Na+ ions. The methodology for this MD simulation was similar to that described for Aβ1–42 in the absence of Cu2+.

Two MD simulations were performed. The first consisted of an evaluation of the conformational change of Aβ1–42 in the α-helix conformation in the presence of 1 Cu2+ ion. The second simulation was performed using Aβ1–42 in the β-sheet conformation (reached by MD simulations using a snapshot retrieve at 13 ns) in the presence of 1 Cu2+ ion to assess the effect of Cu2+ in shaping the secondary structure of Aβ1–42 using the procedure described above.

MD simulation of the binding of galanthamine to Aβ1–42 in the α-helix conformation

The input galanthamine coordinates were taken from the structure with PDB code 1DX6[40] and placed near Hys 6, 13, and 14. Hydrogen atoms were added with the psfgen command within the VMD program,[33] and the structure was minimized using the steepest descent algorithm for 2000 steps using the CHARMM27 force field. The structure was immersed in water (10 Å TIP3 water model), and the charge was neutralized using 2 Na+ ions. The parameters and topology for galanthamine were calculated using the online SwissParam server (http://www.swissparam.ch/). The methodology for the MD simulations was similar to that previously described.

MD simulation analyses

The analyses of the MD simulation results were performed using the carma software package.[38] The root-mean-square deviation (RMSD) of the alpha carbon (Cα) from the backbone residues, the root-mean-square fluctuation (RMSF) of the backbone residue atoms and the ratio of gyration (Rg) of Cα were determined. All the molecular graphical presentations from the docking and MD simulations were created using the PyMOL package (http://www.pymol.org/).

Docking method

A docking study was performed using two different Aβ monomers to construct their respective oligomers. The first monomer was the conformer obtained from the MD simulations at 20 ns using the Aβ PDB code 1Z0Q model 1, and the second was the Aβ obtained from the 2BEG structure. The docking was performed on a ClusPro2.0 server (http://cluspro.bu.edu).[19] To obtain an Aβ dimer (lowest free energy), two of the same monomers were used. The ClusPro2.0 server, one of the top performers at CAPRI (Critical Assessment of Predicted Interactions) round 13–19,[20] was used to predict the possible structure of the oligomers. We submitted to ClusPro two monomers to obtain the dimer, which was used to dock with another monomer to form a trimer, and so forth; this process was performed until an Aβ pentamer was formed.

ClusPro selects the 1000 best scoring solutions and then clusters them according to root mean square deviation (RMSD) considerations. Each cluster is characterized by its number of members, the ClusPro score of the center of the cluster and the lowest ClusPro score found in the cluster. We used the balance favored by hydrophobic interactions, VdW interactions and electrostatic interactions according to ClusPro score.[20]

Evaluation of Aβ142 aggregation by thioflavin T fluorescence

Lyophilized wild-type human Aβ1–42 peptide (chloride salt; Calbiochem) was purchased from Merck Mexico. HEPES sodium salt (>99.5% purity), ThT, CuCl2, and galanthamine were obtained from Sigma-Aldrich (Mexico). All solutions were prepared in fresh Milli-Q water. A freshly prepared solution of Aβ1–42 with a concentration of 62.9 µM was incubated alone and in the presence of 20 µM CuCl2 [22] or 100 µM galanthamine[13] because at these concentrations Aβ1–42 oligomerization is inhibited. The samples were incubated at 37°C in a quartz cell for 24 h with a 0.5-cm path length, and were subsequently stirred at 250 rpm. The increase in ThT fluorescence was measured at 480 nm using 445-nm excitation on a Perkin Elmer LS-55 fluorescence spectrophotometer equipped with a water-jacketed cell holder for temperature control. All the experiments were performed at 37°C using quartz cells with a path length of 1.0 cm, 20 mM HEPES, 100 mM NaCl, a pH of 7.4, and a ThT concentration of 3.3 µM, as described in a previous report.[22] To obtain the corresponding graphics, the results of the spectra for Cu2+ and galanthamine were subtracted from the spectra from ThT.

Circular Dichroism measurements

The lyophilized wild-type human Aβ1–42 peptide was diluted in Milli-Q water to a final concentration of 50 µM, as reported by Salomon.[41] The Aβ1–42 solution was incubated at 37°C in the absence and presence of 20 µM CuCl2 or 100 µM galanthamine. CD spectra were acquired using a JASCO J-815 spectropolarimeter (Jasco, Easton, MD) equipped with a PFD-425S Peltier-type cell holder for temperature control at 37°C and magnetic stirring. Three accumulations of the CD spectra were recorded from 200 to 250 nm using 1.0-mm path-length quartz cells. The data were collected each hour for 3 h of incubation. The data were corrected by subtracting the spectrum of a sample that contained all of the components except Aβ. The data were converted to mean residue ellipticity and analyzed using the CDPro software package.[42]

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Conclusions
  7. Methods
  8. References