Conformational transition of amyloid -peptide - Proceedings of the ...

6 downloads 38 Views 591KB Size Report
Apr 12, 2005 - Communicated by William N. Lipscomb, Harvard University, Cambridge, .... mutated into Water. B3. 300. 50. A25A29A33A37. B4. 400. 50. III. C1 ..... Daura, X., Jaun, B., Seebach, D., van Gunsteren, W. F. & Mark, A. E. (1998).
Conformational transition of amyloid ␤-peptide Yechun Xu*†, Jianhua Shen*†, Xiaomin Luo*, Weiliang Zhu*, Kaixian Chen*, Jianpeng Ma*‡§¶, and Hualiang Jiang*¶储 *Center for Drug Discovery and Design, State Key Laboratory of Drug Research, Shanghai Institute of Materia Medica, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, and Graduate School of Chinese Academy of Sciences, 555 Zuchongzhi Road, Shanghai 201203, China; 储School of Pharmacy, East China University of Science and Technology, Shanghai 200237, China; ‡Department of Biochemistry and Molecular Biology, Baylor College of Medicine, One Baylor Plaza, BCM-125, Houston, TX 77030; and §Department of Bioengineering, Rice University, Houston, TX 77005

The amyloid ␤-peptides (A␤s), containing 39 – 43 residues, are the key protein components of amyloid deposits in Alzheimer’s disease. To structurally characterize the dynamic behavior of A␤40, 12 independent long-time molecular dynamics (MD) simulations for a total of 850 ns were performed on both the wide-type peptide and its mutant in both aqueous solution and a biomembrane environment. In aqueous solution, an ␣-helix to ␤-sheet conformational transition for A␤40 was observed, and an entire unfolding process from helix to coil was traced by MD simulation. Structures with ␤-sheet components were observed as intermediates in the unfolding pathway of A␤40. Four glycines (G25, G29, G33, and G37) are important for A␤40 to form ␤-sheet in aqueous solution; mutations of these glycines to alanines almost abolished the ␤-sheet formation and increased the content of the helix component. In the dipalmitoyl phosphatidylcholine (DPPC) bilayer, the major secondary structure of A␤40 is a helix; however, the peptide tends to exit the membrane environment and lie down on the surface of the bilayer. The dynamic feature revealed by our MD simulations rationalized several experimental observations for A␤40 aggregation and amyloid fibril formation. The results of MD simulations are beneficial to understanding the mechanism of amyloid formation and designing the compounds for inhibiting the aggregation of A␤ and amyloid fibril formation. molecular dynamics simulation

A

lzheimer’s disease (AD), a neurodegenerative disorder, is pathologically characterized by the presence of extracellular senile plaques and intracellular neurofibrillary tangles in the brain (1). The major components of the plaques are amyloid ␤-peptides (A␤s) consisting of 39–43 residues that are proteolytically derived from the widely distributed transmembrane amyloid precursor glycoprotein (APP) (2–4). The amyloid hypothesis suggests that misfolding of A␤ leads to its dysfunctions and fibrillization; the latter is associated with a cascade of neuropathogenetic events to produce the cognitive and behavioral decline hallmarks of AD (4–6). Recently, however, compelling evidence has emerged that not the fibrillar aggregates but the dominant ␤-sheet structure of oligomers and fibril intermediates (protofibrils) are neurotoxic and might be the determinant pathogenic factor in AD (7–9). For example, Walsh et al. (8) demonstrated that cerebral microinjection of cell medium containing abundant A␤ monomers and oligomers but no amyloid fibrils markedly inhibited hippocampal long-term potentiation in rats in vivo. The conflict between the amyloid hypothesis and these new emerging experimental observations has stimulated more and more researchers to study the earliest phases of A␤ assembly and to explore the intrinsic properties of A␤s and the conformational dynamic behaviors of the A␤ monomer (9–15). In addition, it has been established experimentally that the major secondary structure adopted by A␤ depends on the environment. The A␤ monomer favors an ␣-helix structure in a membrane or membrane-mimicking environment such as ionic detergents (15–19). In contrast, A␤ exists mainly as a random coil with few components of ␣-helix and兾or ␤-sheet conformations in aqueous solution (14, 18, 20–22). However, x-ray diffraction measurements on oriented fibril bundles have indicated www.pnas.org兾cgi兾doi兾10.1073兾pnas.0501218102

an extended ␤-sheet structure for Alzheimer’s ␤-amyloid fibrils, with peptide chains in ␤-strand conformations running approximately perpendicular to the fibril axis and hydrogen bonds between peptide chains in each layer occurring roughly parallel to the fiber axis (23, 24). The dominant structure of A␤ in membrane or membrane-like mediums, aqueous solution, and fibrils leads to a postulation that a conversion from an ␣-helix or random coil to a ␤-sheet conformation occurs during (or before) the aggregation of A␤ (25–28). Similar conformational conversions have also been found in proteins of other diseases (29–31). Thus, a rigorous mechanistic study of such conformational transition is critical to understanding and development of therapeutic approaches for the conformational disorders in general, and AD in particular. Unfortunately, because of the fibril’s extreme insolubility and the monomer’s high propensity to aggregate, investigation of the above conformational conversions of A␤ in the atomic detail by means of experimental methods is still intractable (32). Computational simulation with its extremely high time resolution and atomic level representation has been increasingly used in understanding the complex conformational features of polypeptides and in predicting structural preferences (33–38). Simulations have been performed by other groups with a view to elucidating the conformational transition and assembly mechanism of A␤ (10, 12, 13, 32, 39–42) However, these studies are based on the modeled structures of A␤, i.e., fragments or synthesized analogs of A␤, rather than the physiologically produced A␤ in the simulations. In addition, the environment was not considered explicitly, and the simulation time was not long enough to discover the significant conformations. Here, we report on the result of conformational transitions of monomer A␤40 addressed by a series of long time molecular dynamics (MD) simulations. We examined the conformational transitions of A␤40 in both aqueous solution and lipid bilayers. By running six MD simulations in aqueous solution, we show that conformational transition from ␣-helix to coil occurs through helix兾␤-sheet mixed conformations, which is consistent with the recent experimental finding by Kirkitadze et al. (11) that helix兾 ␤-sheet mixed conformations are possible intermediates for A␤ oligomerization. Remarkably, these MD simulations addressed that G25, G29, G33, and G37 play essential roles for A␤40 ␤-sheet formation, and a segment consisting of residues 24–40 may be the core for A␤ oligomerization. This finding was confirmed by running another four MD simulations on the mutant of these glycines. By running an additional two MD simulations, we demonstrate that A␤40 adopts helix as its major secondary structure in the lipid bilayer. However, A␤40 has a tendency to move out of the lipid environment toward the surface lipid Abbreviations: AD, Alzheimer’s disease; A␤, amyloid ␤-peptide; A␤40, residues 1– 40 of wild-type amyloid ␤-peptide; MA␤40, residues 1– 40 of mutant amyloid ␤-peptide; MD, molecular dynamics; DSSP, defined secondary structure of proteins; APP, amyloid precursor glycoprotein; DPPC, dipalmitoyl phosphatidylcholine. †Y.X.

and J.S. contributed equally to this work.

¶To

whom correspondence may be addressed. E-mail: [email protected] or [email protected].

© 2005 by The National Academy of Sciences of the USA

PNAS 兩 April 12, 2005 兩 vol. 102 兩 no. 15 兩 5403–5407

BIOPHYSICS

Communicated by William N. Lipscomb, Harvard University, Cambridge, MA, February 17, 2005 (received for review December 21, 2004)

bilayer. These observations are beneficial to understanding the mechanism of amyloid formation and designing compounds for inhibiting the aggregation of A␤.

Table 1. The detailed information for 12 MD simulations Models Trajectory Temperature, K Time, ns I

A1 A2 A3 A4 A5 A6

300 300 300 300 300 400

150 50 50 50 50 100

II

B1 B2 B3 B4

300 300 300 400

50 50 50 50

III

C1 C2

323 323

100 100

Methods Simulation Systems. Initial coordinates for the wild-type A␤40

with a sequence of DAEFRHDSGYEVHHQKLVFFAEDVG25SNKG29AIIG33LMVG37GVV was taken from the NMR structures determined in aqueous SDS micelles at pH 5.1 (15) (PDB entry 1BA4). For the simulations of A␤40 (or mutated A␤40) in aqueous solution, the peptide was first put into a suitably sized box, of which the minimal distance from the peptide to the box wall was 1.5 nm. Then the box was solvated with the simple point charge (SPC) water model (43). The peptide兾water system was submitted to energy minimization. Afterward, counterions were added to the system to provide a neutral simulation system. The whole system was subsequently minimized again. To set up the A␤40兾dipalmitoyl phosphatidylcholine (DPPC) bilayer, one lipid molecule was first removed from a preequilibrated DPPC bilayer to fit in the C-terminal fragment of the peptide. The peptide兾lipids system was subsequently solvated with the SPC water model. The resulting system was submitted to energy minimization to remove unfavorable contacts and equilibrated for 500 ps with positional restraints on the peptide atoms. Counterions were subsequently added to compensate for the net positive charge of the system. Finally, the whole system was minimized again. MD Simulations. MD simulations were carried out by using the

package (44, 45) with constant number, pressure, and temperature (NPT) and periodic boundary conditions. The GROMACS force field (46) was applied for the peptide, and the lipid parameters adopted were those used in previous MD studies of lipid bilayers (47–50). The linear constraint solver (LINCS) method (51) was used to constrain bond lengths, allowing an integration step of 2 fs. Electrostatic interactions were calculated with the Particle-Mesh Ewald algorithm (52, 53). A constant pressure of 1 bar was applied with a coupling constant of 1.0 ps (54). For the peptide兾DPPC bilayer system the pressure was performed independently in the x, y, and z directions of the whole system. Water, lipids, and peptide were coupled separately to a temperature bath, 300 K for peptide兾 water systems and 323 K for peptide兾DPPC bilayer systems, by using a coupling time of 0.1 ps (54). MD simulations were run on a 128-CPU Silicon Graphics (Mountainview, CA) Origin 3800 server. Analyses were performed by using facilities within the GROMACS package. Secondary structure analyses were carried out employing the defined secondary structure of proteins (DSSP) method (55). Structural diagrams were prepared by using MOLSCRIPT (56) and then rendered with RASTER3D (57). GROMACS

Results To probe how the environment and the sequence specificity affect the conformational transitions of A␤, three models for MD simulations were designed. To make certain the conformational transition is the intrinsic character of A␤ rather than a stochastic output of simulations, several MD simulations were carried out for each system under different initial conditions by assigning different initial velocities on each atom of the simulation systems. In model I, the wild-type A␤40 was solvated with water molecules. Six simulations were conducted for this model: five simulations (A1–A5) adopted different initial velocities assigned on each atom; one simulation (A6) was performed under a higher temperature (400 K). In model II, four glycines (G25, G29, G33, and G37) of A␤40 were mutated by alanines, and then the A␤40 mutant (MA␤40) was solvated with water molecules. Four simulations were performed on this model: three 5404 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0501218102

Mutation

Solvent

Water

G25G29G33G37 mutated into Water A25A29A33A37

DPPC

simulations (B1–B3) were conducted with different initial velocities assigned on each atom; one simulation (B4) was performed at a higher temperature (400 K). In model III, A␤40 was embedded in the environment of the DPPC bilayer, and two simulations (C1 and C2) were performed in two different sizes of the DPPC bilayers (127 and 63 DPPC lipids). In total, 12 MD simulations were performed on A␤40 and MA␤40. More detailed information for these MD simulations is listed in Table 1. Conformational Transition of A␤40 in Aqueous Solution. Six simulations for model I were performed to explore the conformational dynamics of wild-type A␤40 in aqueous solution. The profile of secondary structures of A␤40 along trajectory A1 is shown in Fig. 1a. Obviously, the initial long ␣-helical structure of the peptide disappeared after several nanoseconds; in particular, the ␣-helix of residues 24–37 was converted into several short ␤-strands connected with turns during the first 10 ns. Afterward, three to four ␤-strands, each comprising two residues, existed throughout the remainder of the simulation. However, the central hydrophobic region (residues 16–21) always kept the helical structure throughout the whole 150-ns simulation, with occasional local deviations from ␣-helix to 5-helix or 3-helix. The remainder of

Fig. 1. Structural analyses of A␤40 in trajectory A1. (a) Secondary structures as function of time for A␤40 in trajectory A1 as calculated by DSSP. The structures were analyzed every 750 ps. (b) The snapshot structures of A␤40 extracted from trajectory A1, in which the secondary structure is according to DSSP analysis. C and N denote C-terminal and N-terminal, respectively.

Xu et al.

Table 2. The statistical helix and ␤-sheet components of A␤40 from 10 MD trajectories Helix,* %

␤-sheet, %

I

A1 A2 A3 A4 A5

II

B1 B2 B3

III

C1 C2

20 25 22 35 28 Avg 26† 53 53 54 Avg 53† 38 44 Avg 41†

14 9 14 11 10 Avg 12† 0 0 0 Avg 0† 0 0 Avg 0†

*The sum of ␣-helix, 3-helix, and 5-helix. †Averaged value of each model.

the peptide, residues 1–14, mostly adopts a random coil structure. The snapshot structures of A␤40 extracted from trajectory A1 (shown in Fig. 1b) visualize the secondary structural transformation of A␤40 described above (Fig. 1a). The transformations of secondary structures of A␤40 along the trajectories of simulations A2–A5 are displayed in Fig. 6, which is published as supporting information on the PNAS web site. The profiles of the secondary structure of simulations A2–A5 is similar to that of simulation A1 (Figs. 1 and 6). These five simulations revealed that, in aqueous solution and room temperature (300 K), A␤40 is statistically composed by ⬇26% helix, ⬇12% ␤-sheet, and ⬇62% coil components (Table 2). This finding is in good agreement with the NMR observations (14, 18, 22), which indicated that, in aqueous solution, the monomer of A␤40 favors random coil structure, but ␣-helix and ␤-sheet conformations still exist. Interestingly, the MD simulations indicate that helix兾␤-strand mixed conformations exist in the trajectory of the A␤40 conformational transition. This result is consistent with the experimental results by Soto et al. (16) and Kirkitadze et al. (11). Notably, all of the five trajectories pointed out that V24G25, K28G29, I32G33, and V36G37 form short ␤-strands, indicating that four glycines (G25, G29, G33, and G37) are important for A␤40 to form ␤-sheet in aqueous solution. In another 100-ns MD simulation (simulation A6), the temperature was increased to 400 K, which accelerated the process of A␤40 structure transition. The result is shown in Fig. 2. Similar to simulations A1–A5, ␤-sheet consisting of three or four ␤-strands near the C terminus of A␤40 (residues 24–37) was observed in this simulation (Fig. 2a). The ␤-sheet lasted to ⬇38 ns, and then a short ␣-helix was formed instead of the ␤-sheet in the following 10 ns. After ⬇50 ns, the secondary structure of residues 24–37, which adopts a ␤-sheet structure in trajectories A1–A5, was destroyed, and the entire A␤40 peptide mainly adopted random coil structure throughout the rest of the simulation time. However, there are a few short-lived ␤-strands that appeared in the different positions of the peptide from 50 ns to 100 ns. Different from simulations A1–A5, the helical structure of residues 16–21 lasted only ⬇13 ns (Fig. 2a). Remarkably, this simulation captured the whole process of A␤40 conformational transition from ␣-helix to random coil through helix兾␤-sheet mixed conformations (Fig. 2). These explored conformational changes provide a significant interpretation for the recent result of the kinetic study of A␤ fibrillogenesis by Kirkitadze et al. (11), suggesting that helix兾␤-sheet mixed conformations are possible intermediates for A␤ oligomerization. Xu et al.

Fig. 2. Structural analyses of A␤40 in trajectory A6. (a) Secondary structures as function of time for A␤40 in trajectory A6 as calculated by DSSP. The structures were analyzed every 500 ps. (b) The snapshot structures of A␤40 extracted from trajectory A6, in which the secondary structure is according to DSSP analysis. C and N denote C-terminal and N-terminal, respectively.

Mutation of Four Glycines. To illustrate the importance of G25, G29,

G33, and G37 in the ␤-sheet formation, four MD simulations (B1–B4) were performed on a mutant of A␤40 (MA␤40, model II), in which the four glycines were mutated to alanines, for alanine is an appropriate helix former in aqueous solution. The simulation results are shown in Figs. 3 and 4. Fig. 3a exhibits the time dependence of secondary structure altered for MA␤40 in trajectory B1, and those in trajectories B2 and B3 are shown in Fig. 6. Unlike the wild-type A␤40, ⬇53% residues of MA␤40 adopt helical structures in aqueous solution (Table 2); the main components of the helical structures are ␣-helix and 5-helix (Fig. 3). It is noteworthy that, throughout the three 50-ns simulations, no ␤-sheet structure was observed for residues 24–37 of the MA␤40. MD simulation (B4) at 400 K for MA␤40 revealed a similar result to that of simulations B1–B3 (Fig. 4): the major component of secondary structure is helix; residues 15–20 and 24–34 adopted helix structures throughout the simulation; long lime durable ␤-sheet was not observed although a few ␤-stands were formed sporadically near the N terminus (Fig. 4a). We even continued the simulation for MA␤40 at 400 K to 100 ns, where

BIOPHYSICS

Trajectory

Models

Fig. 3. Structural analyses of MA␤40 in trajectory B1. (a) Secondary structures as function of time for MA␤40 in trajectory B1 as calculated by DSSP. The structures were analyzed every 500 ps. (b) The snapshot structures of MA␤40 extracted from trajectory B1, in which the secondary structure is according to DSSP analysis. C and N denote C-terminal and N-terminal, respectively. PNAS 兩 April 12, 2005 兩 vol. 102 兩 no. 15 兩 5405

Fig. 4. Structural analyses of MA␤40 in trajectory B4. (a) Secondary structures as function of time for MA␤40 in trajectory B4 as calculated by DSSP. The structures were analyzed every 500 ps. (b) The snapshot structures of MA␤40 extracted from trajectory B4, in which the secondary structure is according to DSSP analysis. C and N denote C-terminal and N-terminal, respectively.

the characterized ␤-sheets formed in the wild-type peptide were not observed (Fig. 6B4). All these results indicated that the mutations stabilized the helix structure of A␤40 and illustrated that the four glycines exerted significant control over the process in the formation of ␤-sheet. Inserting A␤40 into DPPC Bilayers. A␤ is a cleavage product of the

single transmembrane protein APP. ␤-secretase cleaves the extracellular domain of APP to generate the N terminus of A␤, and then ␥-secretase performs an unusual proteolysis in the middle of the transmembrane domain of APP to produce the C terminus of A␤ (58). Residues 29–40 of A␤40 are from the transmembrane domain of APP. This special secreted pathway of A␤ raises several questions. What happens to A␤ at the moment when it is completely cut from APP? How does A␤40 undergo conformational transition in the membrane environment? To answer these questions, model III was designed to examine the conformational dynamics of A␤40 in the membrane through inserting residues 29–40 of A␤40 into a DPPC lipid bilayer. Two simulations (C1 and C2) were performed on model III at 300 K. Different sizes of the DPPC bilayer were applied for these two simulations, respectively: 127 DPPC lipids for simulation C1 and 63 DPPC lipids for simulation C2. The profile of secondary structure change of A␤40 (DSSP plot) and several snapshot structures extracted from trajectory C1 are shown in Fig. 5. The DSSP plot illustrates that residues 15–35 kept ␣-helical structure during the 100-ns simulation except that residues 32–35 were once switched to turn structure from 14 to 85 ns but recover to helical structure after 85 ns. The rest of the residues of A␤40 other than residues 15–35 mainly adopted coil or bend structures. During the 100-ns MD simulation in the DPPC bilayer, ␤-sheet structures were not observed, helical structure dominated the secondary structure, and ⬇41% of the residues adopted helical conformation (Table 2). In addition, snapshot structures demonstrate that, during the simulation, A␤40 moves to the interface between DPPC lipids and water molecules, and, after ⬇10 ns, part of A␤40 lies down on the surface of the lipid bilayer. As the simulation goes on, A␤40 tends to totally exit the lipid bilayer (Fig. 5). Similar results were obtained from simulation C2 (Fig. 6). 5406 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0501218102

Fig. 5. Structural analyses of A␤40 in trajectory C1. (a) Secondary structures as function of time for A␤40 in trajectory C1 as calculated by DSSP. The structures were analyzed every 500 ps. (b) The snapshot structures of A␤40 extracted from trajectory C1, in which the peptide is drawn as ribbon, phosphate atoms of lipid are shown as brown balls, the other atoms of the lipid are labeled as sticks, water molecules are displayed as balls and sticks, and the front half of the bilayer is not shown for the sake of clarity.

Discussion The conformational transitions of A␤40 associated with A␤ aggregation and neurotoxicity were studied by using MD simulation method. Detailed analyses were performed for the conformational transition of A␤40 based on 12 MD trajectories. The results demonstrate that A␤40 adopts different conformations in aqueous solution and in the DPPC bilayer, which are in agreement with the experimental results (14–22). Six MD simulations (A1–A6) with different initial velocities assigned to the atoms and different temperatures consistently produced the same result that conformational transition of A␤40 from ␣-helix to ␤-sheet occurs in aqueous solution (Figs. 1 and 2). Our MD simulations captured the whole process of A␤40 conformational transition from ␣-helix to coil structure through an intermediate ␤-sheet structure (Fig. 2). In addition, the MD simulations revealed that a segment of residues 24–37 is the core of A␤40 to form ␤-sheet. In particular, four glycines (G25, G29, G33, and G37), which normally destabilize helical structure, are essential to the ␤-sheet formation. Substitution of these four glycines of A␤40 by alanines increases the stability of the helical structure. ␤-sheets were not observed for residues 24–37 in another three MD simulations (B1–B3) on the A␤40 mutant (MA␤40), and the dominant secondary structure of MA␤40 remained as helix (Figs. 3 and 4). This finding indicates that the sequence of A␤40 is specific to forming ␤-sheet in aqueous solution. This conclusion is in agreement with the opinion of de la Paz and Serrano (59), who found that amyloid fibril formation depends on the peptide sequences based on positional scanning mutagenesis on a designed amyloid peptide. In the DPPC bilayer, helix dominates the secondary structure of A␤40, and no ␤-strand was observed in the two 100-ns MD simulations (C1 and C2) (Fig. 5a and Table 2). This result agrees well with the NMR determination of A␤40 in membranemimicking environment (15, 17, 19). In addition, the segment of A␤40 outside the membrane (residues 1–28) bends toward the surface of the DPPC bilayer, forming electrostatic and van der Xu et al.

Waals interactions to the polar heads of the lipids (Fig. 7, which is published as supporting information on the PNAS web site). These interactions facilitate A␤40 to completely leave the lipids, moving toward the surface of the DPPC bilayer (Fig. 5b). Even at the interface between aqueous solution and the lipid bilayer, A␤40 still keeps helix as its dominant secondary structure. Abundant experimental data indicated that the major component of the plaques in AD patient’s brain is ␤-peptides, which aggregate into amyloid fibrils with cross-␤-structure (2, 23, 24). An essential question concerning the plaques is the origin of the amyloid fibrils formation. Taken together, 12 long-time MD simulations provide a possible mechanism of amyloid fibrils aggregation. Cleaved by ␤- and ␥-secretases, A␤ peptides were released from APP (58). Firstly, residues 29–40 of A␤40 are embedded into the cellular membrane; this environment is beneficial to stabilizing the helical structure of A␤ (Fig. 5a). Our MD simulations (C1 and C2) indicate that A␤ tends to leave the pure DPPC lipid bilayer (Fig. 5b). In fact, the concentration of cholesterol in the cytofacial leaflet of brain synaptic plasma membrane is lower in aged people (60), and A␤ cannot insert into the membrane containing less cholesterol (61). Once A␤ exits the membrane and enters into the extracellular solution, the

segment of residues 29–40 has a high tendency to form short ␤-sheets (Figs. 1 and 2). Accordingly, we suggest that the short ␤-sheet formation of residues 29–40 is a potential driving force for A␤s aggregating into amyloid fibrils, implying that small molecules or peptides that can bind tightly with this segment may be developed as inhibitory therapeutic agents. Although it is clear from our MD simulations that amyloid fibrils formation of A␤40 is associated with its dynamic properties, a complete description of the mechanism of how A␤40 monomers aggregate together requires further analysis using both experimental and theoretical methods.

1. Selkoe, D. J. (1991) Neuron 6, 487–498. 2. Miller, D. L., Papayannopoulos, I. A., Styles, J., Bobin, S. A., Lin, Y. Y., Biemann, K. & Iqbal, K. (1993) Arch. Biochem. Biophys. 301, 41–52. 3. Kang, J., Lemaire, H. G., Unterbeck, A., Salbaum, J. M., Masters, C. L., Grzeschik, K. H., Multhaup, G., Beyreuthe, K. & Muller-Hill, B. (1987) Nature 325, 733–736. 4. Mattson, M. P. (1997) Physiol. Rev. 77, 1081–1132. 5. Selkoe, D. J. (2001) Physiol. Rev. 81, 741–766. 6. Hardy, J. & Selkoe, D. J. (2002) Science 297, 353–356. 7. McLean, C. A., Cherny, R. A., Fraser, F. W., Fuller, S. J., Smith, M. J., Beyreuther, K., Bush, A. I. & Masters, C. L. (1999) Ann. Neurol. 46, 860–866. 8. Walsh, D. M., Klyubin, I., Fadeeva, J. V., Cullen, W. K., Anwyl, R., Wolfe, M. S., Rowan, M. J. & Selkoe, D. J. (2002) Nature 416, 535–539. 9. Rosenblum, W. I. (2002) Neurobiol. Aging 23, 225–230. 10. Hwang, W., Zhang, S. G., Kamm, R. D. & Karplus, M. (2004) Proc. Natl. Acad. Sci. USA 101, 12916–12921. 11. Kirkitadze, M. D., Condron, M. M. & Teplow, D. B. (2001) J. Mol. Biol. 312, 1103–1119. 12. Straub, J. E., Guevara, J., Huo, S. H. & Lee, J. P. (2002) Acc. Chem. Res. 35, 473–481. 13. MA, B. Y. & Nussinov, R. (2002) Proc. Natl. Acad. Sci. USA 99, 14126–14131. 14. Zhang, S., Iwata, K., Lachenmann, M. J., Peng, J. W., Li, S., Stimson, E. R., Lu, Y. A., Felix, A. M., Maggio, J. E. & Lee, J. P. (2000) J. Struct. Biol. 130, 130–141. 15. Coles, M., Bicknell, W., Watson, A. A., Fairlie, D. P. & Craik, D. J. (1998) Biochemistry 37, 11064–11077. 16. Soto, C., Castano, E. M., Frangione, B. & Inestrosa, N. C. (1995) J. Biol. Chem. 270, 3063–3067. 17. Sticht, H., Bayer, P., Willbold, D., Dames, S., Hilbich, C., Beyreuthe, K., Frank, R. W. & Rosch, P. (1995) Eur. J. Biochem. 233, 293–298. 18. Barrow, C. J., Yasuda, A., Kenny, P. T. M. & Zagorski, M. G. (1992) J. Mol. Biol. 225, 1075–1093. 19. Shao, H. Y., Jao, S. C., Ma, K. & Zagorski, M. G. (1999) J. Mol. Biol. 285, 755–773. 20. Serpell, L. (2000) Biochim. Biophys. Acta 1502, 16–30. 21. Good, T. A. & Murphy, R. M. (1995) Biochem. Biophys. Res. Commun. 207, 209–215. 22. Barrow, C. J. & Zagorski, M. G. (1991) Science 253, 179–182. 23. Blake, C. & Serpell, L. (1996) Structure 4, 989–998. 24. Sunde, M., Serpell, L. C., Bartlam, M., Fraser, P. E., Pepys, M. B. & Blake, C. C. F. (1997) J. Mol. Biol. 273, 729–739. 25. Teplow, D. B. (1998) Amyloid 5, 121–142. 26. Pike, C. J., Walencewicz-Wasserman, A. J., Kosmoski, J., Cribbs, D. H., Glabe, C. G. & Cotman, C. W. (1995) J. Neurochem. 64, 253–265. 27. Walter, M. F., Mason, P. E. & Mason, R. P. (1997) Biochem. Biophys. Res. Commun. 233, 760–764. 28. Lee, J. P., Stimson, E. R., Ghilardi, J. R., Mantyh, P. W., Lu, Y. A., Felix, A. M., Llanos, W., Behbin, A., Cummings, M., Van Criekinge, M., et al. (1995) Biochemistry 34, 5191–5200. 29. Pan, K., Baldwin, M., Nguyen, J., Gasset, M., Serban, A., Groth, D., Mehlhorn, I., Huang, Z., Fletterick, R. J., Cohen, F. E. & Prusiner, S. B. (1993) Proc. Natl. Acad. Sci. USA 90, 10962–10966.

30. Weinreb, P. H., Zhen, W., Poon, A. W., Conway, K. A. & Lansbury, P. T., Jr. (1996) Biochemistry 35, 13709–13715. 31. Conway, K. A., Harper, J. D. & Lansbury, P. T., Jr. (1998) Nat. Med. 4, 1318–1320. 32. Mager, P. P., Penke, B., Walter, R., Harkany, T. & Hartig, W. (2002) Curr. Med. Chem. 9, 1763–1780. 33. Daura, X., Jaun, B., Seebach, D., van Gunsteren, W. F. & Mark, A. E. (1998) J. Mol. Biol. 280, 925–932. 34. Duan, Y. & Kollman, P. A. (1998) Science 282, 740–744. 35. Simmerling, C., Strockbine, B. & Roitberg, A. E. (2002) J. Am. Chem. Soc. 124, 11258–11259. 36. Snow, C. D., Nguyen, H., Pande, V. S. & Gruebele, M. (2002) Nature 420, 102–106. 37. Mayor, U., Guydosh, N. R., Johnson, C. M., Grossmann, J. G., Sato, S., Jas, G. S., Freund, S. M. V., Alonso, D. O. V., Daggett, V. & Fersht, A. R. (2003) Nature 421, 863–867. 38. Zangi, R., de Vocht, M. L., Robillard, G. T. & Mark, A. E. (2002) Biophys. J. 83, 112–124. 39. Massi, F. & Straub, J. E. (2001) Biophys. J. 81, 697–709. 40. Mager, P. P. (1998) Med. Res. Rev. 18, 403–430. 41. Klimov, D. K. & Thirumalai, D. (2003) Structure 11, 295–307. 42. Zanuy, D., Ma, B. Y. & Nussinov, R. (2003) Biophys. J. 84, 1884–1894. 43. Berendsen, H. J. C., Postma, J. P. M., van Gunsteren, W. F. & Hermans, J. (1981) in Intermolecular Forces, ed. Pullman, B. (Reidel, Dordrecht, The Netherlands), pp. 331–342. 44. Berendsen, H. J. C., van der Spoel, D. & van Drunen, R. (1995) Comput. Phys. Commun. 91, 43–56. 45. Lindahl, E., Hess, B. & van der Spoel, D. (2001) J. Mol. Model. 7, 306–317. 46. van Gunsteren, W. F. & Berendsen, H. J. C. (1987) Gromos-87 Manual (Biomos, Groningen, The Netherlands). 47. Tieleman, D. P., Sansom, M. S. P. & Berendsen, H. J. C. (1999) Biophys. J. 76, 40–49. 48. Berger, O., Edholm, O. & Jahnig, F. (1997) Biophys. J. 72, 2002–2013. 49. Marrink, S. J., Berger, O., Tieleman, P. & Jahnig, F. (1998) Biophys. J. 74, 931–943. 50. Tieleman, D. P., Berendsen, H. J. C. & Sansom, M. S. P. (1999) Biophys. J. 76, 1757–1769. 51. Hess, B., Bekker, H., Berendsen, H. J. C. & Fraaije, J. G. E. M. (1997) J. Comp. Chem. 18, 1463–1472. 52. Darden, T., York, D. & Pedersen, L. (1993) J. Chem. Phys. 98, 10089–10092. 53. Essmann, U., Perera, L., Berkowitz, M. L., Darden, T., Lee, H. & Pedersen, L. G. (1995) J. Chem. Phys. 103, 8577–8592. 54. Berendsen, H. J. C., Postma, J. P. M., van Gunsteren, W. F., DiNola, A. & Haak, J. R. (1984) J. Chem. Phys. 81, 3684–3690. 55. Kabsch, W. & Sander, C. (1983) Biopolymers 22, 2577–2637. 56. Kraulis, P. J. (1991) J. Appl. Crystallogr. 24, 946–950. 57. Merrit, E. A. & Murphy, M. E. P. (1994) Acta Crystallogr. D50, 869–873. 58. Esler, W. P. & Wolfe, M. S. (2001) Science 293, 1449–1454. 59. de la Paz, M. L. & Serrano, L. (2004) Proc. Natl. Acad. Sci. USA 101, 87–92. 60. Igbavboa, U., Avdulov, N. A., Schroeder, F. & Wood, W. G. (1996) J. Neurochem. 66, 1717–1725. 61. Ji, S. R., Wu, Y. & Sui, S. F. (2002) J. Biol. Chem. 277, 6273–6279.

Xu et al.

PNAS 兩 April 12, 2005 兩 vol. 102 兩 no. 15 兩 5407

BIOPHYSICS

This work was supported by State Key Program of Basic Research of China Grant 2002CB512802; the Key Project for New Drug Research from the Chinese Academy of Sciences; 863 Hi-Tech Program Grants 2002AA233061, 2001AA235051, 2001AA235041, 2002AA104270, 2002AA233011, and 2003AA235030; National Natural Science Foundation of China Grants 20372069, 29725203, and 20072042; and Shanghai Science and Technology Commission Grant 02DJ14006. J.M. was partially supported by National Science Foundation CAREER Award MCB-0237796 (U.S.A.); he is also a recipient of the Award for Distinguished Young Scholars Abroad from the National Natural Science Foundation of China.