molecular dynamics studies of hexamers of amyloid-β peptide (16–35) and its mutants: influence of...

Molecular Dynamics Studies of Hexamers of Amyloid-bPeptide (16–35) and Its Mutants: Influence of ChargeStates on Amyloid Formation

Wei Han1 and Yun-Dong Wu1,2*1Department of Chemistry, The Hong Kong University of Science and Technology, Kowloon, Hong Kong, China2State Key Lab of Molecular Dynamics and Stable Structures, College of Chemistry, Peking University, Beijing, China

ABSTRACT To study the early stage of amy-loid-b peptide (Ab) aggregation, hexamers of thewild-type (WT) Ab16–35 and its mutants with amy-loid-like conformations have been studied by mo-lecular dynamics simulations in explicit water fora total time of 1.7 ms. We found that the amyloid-like structures in the WT oligomers are destabi-lized by the solvation of ionic D23/K28 residues,which are buried in the fibrils. This means that thedesolvation of D23/K28 residues may contribute tothe kinetic barrier of aggregation in the earlystage. In the E22Q/D23N, D23N/K28Q, and E22Q/D23N/K28Q mutants, hydration becomes much lesssignificant because the mutated residues have neu-tral amide side-chains. These amide side-chains canform linear cross-strand hydrogen bond chains, or‘‘polar zippers’’, if dehydrated. These ‘‘polar zip-pers’’ increase the stability of the amyloid-like con-formation, reducing the barrier for the early-stageoligomerization. This is in accord with experimen-tal observations that both the D23/K28 lactamiza-tion and the E22Q/D23N mutation promote aggrega-tion. We also found that the E22Q/D23N mutant pre-fers an amyloid-like conformation that differs fromthe one found for WT Ab. This suggests that differ-ent amyloid structures may be formed under differ-ent conditions. Proteins 2007;66:575–587. VVC 2006

Wiley-Liss, Inc.

Key words: molecular dynamics; amyloid-b pep-tide; oligomer; mutation; Alzheimer’sdisease

INTRODUCTION

Alzheimer’s disease (AD) is a neurodegenerative disease,characterized by the formation of cerebral amyloid pla-ques.1,2 Such fibrillar plaques, composed of aggregates ofamyloid-b (Ab) peptide of 39–43 amino acids,3 have beenshown to be correlated with the pathogenesis of AD.4,5 Ithas been suggested that the inhibition of toxic amyloid6,7

formation may benefit AD therapies.8 Thus, it is importantto understand the plaque-forming mechanism of Ab pepti-des, particularly the structural features of all states duringthis process.The formation of fibrils is a complicated and multi-stage

process. Starting from a soluble monomeric state, fibrils

may go through possible intermediate stages such as be-coming soluble oligomers9,10 and protofibrils [Fig. 1(a)].11,12

The amyloid conformation of Ab peptides in fibrils hasbeen well characterized and shown to possess in-registerparallel b-sheets.13 Tycko and coworkers showed thatmonomers in the fibrils have bends in the E22-G29 re-gion14 to bring b-sheets V12-A21 and A30-A40 into hydro-phobic contact15 and that the ionic side-chains of D23 andK28 form buried salt bridges [Fig. 1(b,c), left].16–18

Recently, much attention has been paid to solubleoligomers and protofibrils since they are thought to beneurotoxic forms of Ab19–21 Soluble oligomers may play animportant role in amyloidogenesis at very early stage.9,10

Ab1–40, Ab1–42, Ab10–35, and Ab10–30 all have soluble low-molecular-weight (LMW) oligomers, including 2–8 mono-mers,9,10,13,22 which are in dynamic equilibrium withmonomers and protofibrils [Fig. 1(a)].9,10,21 Morphologicalstudies have indicated that LMW oligomers are probab-ly the building units to initiate or elongate protofibrils[Fig. 1(a)].21–27 Because Ab structures in protofibrilsresemble those in fibrils in many respects, such as second-ary structures (strand-loop-strand, SLS)21 and importantregions involving in cross-b hydrogen bonds (HB),28 thesesoluble oligomers may possess amyloid-like conformations.Recently, we found through long-time molecular dynamics(MD) simulations on Ab10–35 that a SLS structure is a gen-eral conformational feature of Ab monomers in solution.29

This SLS structure, which resembles that of monomers infibrils, was found both in the wild-type (WT) and theE22Q/D23N mutant. A recent experiment using high-reso-lution atomic force microscopy technique also indicatedthat a monomeric Ab deposited on a mica surface adoptsan SLS conformation, which can assemble into hydratedamyloid-like LMW oligomers that further associate intoprotofibrils.27 Therefore, the stability of amyloid-like solu-ble oligomers may be directly related to the rate of protofi-bril formation in the early stage of aggregation.

Grant sponsor: RGCHK; Grant number: HKUST6083/02M; Grantsponsor: NSFC; Grant number: 20225312.

*Correspondence to: Yun-Dong Wu, Department of Chemistry, TheHong Kong University of Science and Technology, Kowloon, HongKong, China. E-mail: [email protected]

Received 1 June 2006; Revised 14 August 2006; Accepted 23 August2006

Published online 17 November 2006 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/prot.21232

VVC 2006 WILEY-LISS, INC.

PROTEINS: Structure, Function, and Bioinformatics 66:575–587 (2007)

The E22Q and/or D23N mutations have been impli-cated in early-onset familial ADs. They are also found topromote the formation of oligomers, protofibrils, or/andfibrils.10,11,30 Neutralization of D23 and K28 has alsobeen suspected to reduce the lag time of aggregation.31

On the other hand, the formation of D23/K28 salt bridgehas been found to be important for fibrils.15–18 However,in the case of the E22Q/D23N double mutation, saltbridge formation becomes impossible. Without the saltbridge formation, the reason for the aggregation promo-tion observed for these neutralization mutations is notapparent. We seek to determine how the charge states ofthese residues influence the stability of soluble oligomers.The polymorphism of fibrils is another important sub-

ject of study. It has been suggested that different amyloidconformations of Ab peptides in fibrils can lead to distinctmorphologies of fibrils32 and the preference for differentamyloid-like conformations of Ab peptides may exist even

in LMW oligomer states.33 An interesting experiment hasshown that fibrils of Ab10–35 formed under acidic andneutral conditions exhibit different morphologies.34 E22and D23 are anionic under neutral conditions but theyare neutral under acidic conditions. It may be possiblethat the polymorphism of Ab amyloid can be induced bydistinct charge states of these residues, or the differentcharge states of these residues may lead to different amy-loid-like conformations in LMW oligomers.

Ab oligomers have been studied extensively in previousMD simulations.17,18,35–43 These simulations mainly focusedon the stabilities of Ab oligomers and their dynamics offormation. The effect of the charge states of amino acidresidues on the conformational stabilities of oligomers hasnot been the focus of MD investigation in the past. In thispaper, we report on a long-time (total 1.7 ls) MD simula-tion on the hexamers of WT Ab16–35 and its mutants,including ED (E22Q, D23N), DK (D23N, K28Q), and EDK(E22Q, D23N, K28Q) [Fig. 1(c)]. These simulations revealthe important roles of charge states of these residues ininfluencing the stability and conformational preference ofAb oligomers.

MATERIALS AND METHODSModel

The model peptide in this study was Ab16–35 (sequence:K16LVFFAEDVGSNK GAIIGLM35). Its C-terminal wascapped with an amide group, similar with experimentalconditions.34 To construct hexamers with flat b-helicalstructures, we referred to the crystal structure of SulfD(PDB entry 1vh4). Its backbone could be best fitted to thatof a crystal structure of GNNQQNY fibrils.44 We choose thebackbones of regions 344–363 and 214–233 of SulfD to con-struct the monomer with ‘‘TopA’’ and ‘‘TopB’’ conformations[Fig. 1(c)], where the difference between TopA and TopBconformations will be discussed in the following section. Sixmonomers stack together with the periodicity of 5 A to formin-register parallel b-sheets. No twist of b-sheets was intro-duced in hexamer models. In addition, we also built up adecamer of Ab14–36 in a similar way.

Simulation Setups

All simulations were performed with the GROMACS3.2.1 software package.45 The GROMOS96 force field wasused for each of the simulations.46 Peptide oligomers wereput into rhombic dodecahedron boxes that contain 8000–9000 single point charge (SPC) water molecules.47 Chlorideions were used to neutralize systems. Bond length was con-strained by the LINCS algorithm,48 and water moleculeswere restrained by the SETTLE algorithm.49 We increasedtime step up to 5 fs by increasing mass of hydrogen to4 a.u. and using dummy atoms. The previous studies havedemonstrated that this cannot perturb thermodynamicsand dynamics of systems.29,50 The twin range cut-off of0.9/1.4 nm was used for nonbonded interactions. Electro-static interactions were treated by the reaction fieldmethod with the dielectric constant of 78.51 The particle

Fig. 1. (a) Scheme of the fibril aggregation process. (b) Front viewsof the amyloid conformation of hexamers. The yellow hashed lines sig-nify in-register HBs. (c) Top views of the found amyloid topology for WT(TopA) and the one proposed here (TopB). [Color figure can be viewedin the online issue, which is available at www.interscience.wiley.com.]

576 W. HAN AND Y.-D. WU

PROTEINS: Structure, Function, and Bioinformatics DOI 10.1002/prot

mesh Ewald (PME) method was used for electrostatics inlong supporting runs.52

All simulations started with a 5000-step optimization,followed by a 400 ps simulation at 100 K. Temperatureand pressure were controlled by an external thermostatand a pressure bath with coupling constants of 0.1 and1.0 ps, respectively.53 The obtained structures were re-optimized and a 200 ps simulation with peptides restrainedwas performed at 340 K. The final structures were the ini-tial points of product runs. We also performed simulationswith OPLS-all atom force field54 and TIP4P solvent model.55

In these simulations, the twin range cut-off was set to0.9/1.2 nm and the time step was set to 2 fs. The set updetails are shown in Table I.

Definition of HB

A HB is defined as follows: the distance between thedonor (N) and the acceptor (O) is shorter than 0.35 nmand the angle (N��H��O) is larger than 1208.

Definition of Hydration

If the O atom of any water molecule is within 0.35 nmof the Od atoms of any Asp, of the Nf atom of any Lys, ofthe Od/Nd atoms of any Asn, or of the Oe/Ne atoms of anyGln, it is considered as a water molecule that hydratesthe sidechain of this amino acid.

Calculation of Twist Angles u

Twist angle y, as defined by Chothia,56 is the dihedralangle between neighboring b-strands in b-sheets. In amy-loid-like oligomers of Ab, both regions 16–21 and 30–35form b-sheets. We only computed the y of the sheet thatwas better kept for each condition. As will be shown later,the sheet of region 16–21 is better than that of region 30–35 in all ED and EDK simulations. y is thus calculated asthe dihedral of (L17i) � (A21i) � (A21iþ1) � (L17iþ1), wherei and i þ 1 are two neighboring monomers. Only carbonylcarbons of backbones were considered to compute y. Simi-larly, the dihedral of (A30i) � (L34i) � (L34iþ1) � (A30iþ1)is used to calculate y of the rest of simulations. The valuesof dihedral angles were averaged over nonterminal pepti-des of oligomers for each frame of simulations.

Calculation of Shape Complementarity Scores

Shape complementarity (Sc) scores measure the Sc oftwo atomic surfaces by comparing the directions of unitvectors normal to the two surfaces, emanating from near-est points on the opposed surfaces.57 The average dot prod-uct of the pair of vectors approaches 1.0 as the two surfacesfollow each other perfectly. All parameters to compute Scare the same as in the literature.57 The Sc scores in thisstudy were calculated with the CCP4 program packages.58

RESULTS AND DISCUSSIONGeneral Strategy

Our general strategy in this study was to constructamyloid-like hexamers of WT Ab16–35 peptides and their

mutants [Fig. 1(b,c)] as described in the Methods section,and then to check whether or not the amyloid-like struc-tures could survive in an aqueous environment at 340 K,the estimated optimal temperature for Ab aggregation,59

through a series of NPT simulations (Table I). The overallintactness of amyloid-like structures is reflected in thetotal number of in-register HBs,13,15 i.e. in the parallelb-sheet HBs between the same residues of neighboringpeptides of oligomers [yellow hashed lines in Fig. 1(b)]. Ina hexamer, there are five in-register HB layers, whichare sandwiched by neighboring peptides [A–F inFig. 1(b)]. In in-register parallel b-sheets of Ab16–35, themaximum number of HBs of each layer is 20. Thus, 100 in-register HBs is the limit for a perfect amyloid structure.The percentages of HBs maintained in the actual simula-tions indicate the extent to which the oligomers wereamyloid-like.

The five layers are not equivalent. Layer I betweenpeptides C and D is positioned the deepest inside the hex-amers. Layer III between peptides A and B or betweenpeptides E and F is the most exposed to the solvent.Layer II (B and C or D and E) is between layers I and III.The three layers may respond differently to solvent per-turbation because of their positions. The analysis of theHBs in these layers reflects the influence of solvation onthe amyloid-like structures. Furthermore, regions 16–21[Fig. 1(b), blue] and 30–35 (red) of Ab are shown to be inb-strands and critical to amyloidogenesis60 while region22–29 is in a bend.14,15 The maximum numbers of HBs inthese regions are 30, 40, and 30, respectively. The num-bers of the HBs in these three regions can reflect theintactness of the amyloid-like structures in these regions.

The percentages of in-register HBs in regions 16–21,22–29, and 30–35 and layers I, II, and III were averagedwithin the first 5 ns, from 5 to 50 ns and after 50 ns foreach kind of simulation with the same mutation and theconformation (TopA or TopB). They are listed in Table II.

TABLE I. The Set Up Details of All MD Simulations

Namea Size (atom) Time (ns)

TopA-WT 25,434 200 3 2; 100 3 2; 50b

TopA-ED 27,138 100 3 2TopA-DK 27,141 100; 50 3 2TopA-EDK 27,123 180; 50TopB-ED 25,623 100; 50 3 2TopA-PME-WTc 25,896 150; 50TopA-DEC-WTd 44,939 45TopA-OPLS-WTe 30,808 16TopA-OPLS-EDK 31,000 16

aTopA-WT are WT simulations with ‘‘TopA’’ conformation. TopB con-formation can be obtained by flipping region N27-M35 of ‘‘TopA’’ con-formation.bTopA-WT had two 200 ns, two 100 ns, and one 50 ns simulationswith distinct initial atomic velocities.cThe PME method was used while other simulations were performedwith the reaction field method.dThe simulation of a decamer of Ab14–36.eOPLS indicates that the simulation was preformed with the OPLS-all-atom force field and TIP4P water model while other simulations wereperformed with the GROMOS96 force field and SPC water model.

577MD SIMULATIONS OF Ab HEXAMERS


Stabilities of the Amyloid-like Conformationsin WT Hexamers

As shown in Table II, the simulation results for the WTusing the PME method52 (TopA-PME WT) and the reac-tion field method51 (TopAWT) are similar, indicating thateither method can be used. In the first 5 ns, the amyloid-like conformation is largely preserved. However, afterthat, the amyloid-like conformation starts to decay, asindicated by the reduced percentage of in-registered HBs.The loss of HBs mainly occurs in the 16–21 and 22–29regions, which drop to about 30–40 and 20–30% in-regis-ter HBs, respectively. The sheet structure in the 30–35region is more stable. The loss of HBs in the HB layersmainly occurs in the external peptide chains (layer III),where the percentage of in-register HBs drops to about15%. That a greater loss of the HBs occurs in the more-solvent exposed parts implies that the amyloid-like con-formation may be somewhat intolerant to solvation inWT LMW oligomers.

Hydration of Amyloid-like Conformationsof WT Hexamers

The D23/K28 charged sidechains of WT amyloid-likeoligomers are buried inside.16–18 The dehydration of suchcharge pairs from the aqueous phase to form buried saltbridge was suspected to be unfavorable.31 On the otherhand, a channel composed of the F19, A21, A30, and I32sidechains has been identified by Tycko and coworker[Fig. 1(b)].16,18 This channel can hold water molecules

that can hydrate the buried charges and provide stabilityto the amyloid fibrils. It would be interesting to deter-mine whether hydration in this channel is enough for thesolvated amyloid-like oligomers and if hydration of buriedcharge groups is related to the gradual loss of amyloid-like structures in WT oligomers.

To this end, we analyzed the hydration water moleculesof the charged groups of the D23/K28 sidechains (seeMaterials and Method). As shown in Table II, the totalnumber of hydration water molecules of the D23/K28pairs in peptides B–E gradually increases from an initialvalue of about 11 to 16–19 in the end.* On the other hand,each D23/K28 pair in peptide A or F, which is solventexposed, has about eight hydration water molecules (notshown in Table II). This value remained unchanged dur-ing the simulations. The further hydration of D23/K28pairs of peptide monomers B–E is thus more likely corre-lated with the structural loss of WToligomers.

*The column of hydration number in Table II records the numbersof water molecules hydrating polar sidechain atoms of residues 23and 28 in TopA or residues 23 and 27 in TopB. These are roughly thenumbers of water molecules inside the channels only if these polarsidechains are buried inside the channel. This is the case for TopA-WT/PME-WT within the first 5 ns, TopA-DK/EDK and TopB-EDthroughout the simulations. For TopA-ED, the channel disruptswithin 5ns due to strong repulsions among charged K28 sidechains.Therefore, the number of water molecules in the channel of TopA-EDis roughly estimated by the average number of water molecules asso-ciated with N23 and K28 within the period of 100–200 ps of simula-tions, which is about 8.8.

TABLE II. The Averaged Results of the Percentages of In-register HBs, the Numbers of Hydration WaterMolecules, and the Probability of Finding Flat b-Sheets

SimulationPeriod(ns)

% of in-register HBs in different regionsaHydrationnumberb

Flatsheet (%)c16–21 22–29 30–35 I II III

TopA PME-WT 0–5 63 � 10 45 � 8 67 � 10 70 � 10 60 � 10 50 � 10 11 � 2 21.45–50 43 � 13 30 � 10 60 � 10 65 � 10 50 � 10 25 � 10 14 � 3 13.3>50 33 � 3 28 � 5 57 � 7 65 � 10 50 � 10 15 � 5 19 � 3 0.0

TopAWT 0–5 63 � 7 55 � 13 77 � 7 75 � 10 70 � 10 55 � 10 11 � 3 43.45–50 43 � 13 38 � 13 70 � 10 65 � 10 55 � 10 30 � 10 14 � 4 12.1>50 40 � 7 20 � 8 60 � 10 60 � 10 50 � 10 15 � 10 16 � 4 1.2

TopA EDK 0–5 80 � 3 80 � 8 70 � 13 80 � 10 80 � 10 75 � 5 1 � 2 4.45–50 80 � 7 83 � 8 70 � 7 80 � 10 80 � 5 75 � 5 1 � 1 10.5>50 77 � 7 78 � 8 63 � 7 75 � 10 75 � 5 70 � 5 1 � 1 1.8

TopA DK 0–5 63 � 10 70 � 10 63 � 7 70 � 10 70 � 10 60 � 10 4 � 2 15.55–50 57 � 10 78 � 10 67 � 7 75 � 10 75 � 10 60 � 10 1 � 1 1.7>50 60 � 3 75 � 5 63 � 7 80 � 5 80 � 5 50 � 5 0 � 1 0.5

TopA ED 0–5 77 � 7 40 � 13 60 � 17 55 � 15 60 � 10 55 � 5 10 � 3 12.75–50 73 � 7 45 � 10 53 � 17 55 � 10 60 � 10 55 � 10 11 � 3 3.3>50 70 � 10 35 � 5 43 � 13 50 � 10 55 � 5 40 � 10 14 � 3 8.7

TopB ED 0–5 80 � 7 73 � 8 80 � 3 80 � 5 80 � 5 75 � 5 2 � 2 59.35–50 80 � 3 70 � 5 80 � 3 80 � 5 75 � 5 75 � 5 1 � 1 71.8>50 83 � 3 70 � 5 80 � 3 80 � 5 80 � 5 75 � 5 1 � 1 75.7

Maximum Values 30 40 30 20 20 20 NDd ND

aNumbers in bold are �70% of the maximum values of the same columns; numbers in italic are �30% of these maxima.bThe total numbers of water molecules that hydrate sidechains of residues 23 and 28 in peptide B–E in the TopA model or hydrate sidechains ofresidues 23 and 27 in peptide B–E with TopB model.cProbability of finding flat sheets with �38 < y < 38.dNot determined.



Inspection of the 150 ns TopA-PME-WT simulationreveals how the hydration can lead to structural loss[Fig. 2(b)]. At 1 ns, the amyloid-like structures are keptin peptides A–F. There are eight water molecules in thechannel (green balls) and only two water molecules out-side (orange balls in Fig. 2). At 20 ns, although the totalnumber of hydration water molecules increases to 15,eight of them are outside of the channel and hydrate thesolvent-exposed sidechains of D23 in peptide B and ofK28 in peptide E. These water molecules break the saltbridge between K28 in peptide E and D23 in peptide Fand tear peptides A and F away from peptides B–E. Thenumber of water molecules in the channel, however, doesnot increase. At 150 ns, more hydration water moleculesoutside the channel cause buried D23/K28 pairs to beexposed to the solvent, leading to further destruction ofthe salt bridges and separation of the neighboring pepti-des. Interestingly, the number of water molecules in the

channel drops to four since the channel itself is partlydestroyed.

Similar analyses of other WT simulations [Fig. 2(c)]indicate that: (1) although hydration in the channel doesoccur, it appears to be inadequate for the soluble oligom-ers; (2) the channel is unable to hold more water mole-cules beyond the limit of about two water molecules perD23/K28 pair, and the buried D23/K28 pairs must beexposed to the solvent; (3) the hydration of buried D23/K28 pairs is the main reason for the destruction of WTamyloid-like oligomers in this study.

Stabilities of the EDK and DK Mutants

Compared with WT, amyloid-like conformations of theEDK (Table II, TopA-EDK) and DK (Table II, TopA-DK)mutants are much more stable, as shown in Figure 3.Even after 50 ns, the EDK and DK hexamers have about

Fig. 2. (a) The top views of WT with the TopA model. (b) The structural changes in a TopA-PME-WT simulation. Sidechains of D23 (red) and K28(blue) are shown. Yellow and cyan arrows indicate the salt bridges and the separation of the neighboring peptides, respectively. The spheres are watermolecules hydrating residues 23 and 28 in peptides B–E. (c) The last frames of all other WT simulations.



60–77% in-register HBs in region 16–21 and about 75–78% in region 22–29. These results were obtained usingthe GROMOS96 force field46 and SPC water model.47 Wealso performed simulations using the OPLS force field54

and the TIP4P solvent model,55 and found that the rela-tive stability of the amyloid-like conformation in WT andEDK oligomers was not changed, as shown in Figure 4.These additional simulations gave us confidence in oursimulation results. As shown in Table II, both the insidelayers (I and II) and the outside layers (III) of thesemutants are more stable than those of WT, as indicated bythe higher percentages of in-register HBs. The hydrationbehaviors of the N23/Q28 residues of the EDK and DKmutants are in sharp contrast to those of the D23/K28pairs of WT. Most water molecules in the channel at thebeginning are quickly squeezed out during all EDK andDK simulations. As shown in Table II, the total number ofhydration water molecules, which is about 11 at the begin-ning of the simulation, is quickly reduced to about 1–4 inthe first 5 ns and is further reduced to about 1 after 50 ns.Therefore, internal solvation is unlikely to occur.Perutz proposed that Asn-rich or Gln-rich proteins form

amyloids in Huntington Disease.61 In amyloid structures,the sidechain amide groups of these amino acids can formextra cross-strand HBs (between two b-strands), called po-lar zippers, to stabilize b-sheets in addition to the backboneHBs. In the in-register parallel b-sheets of EDK and DK,the sidechains of the same residues are stacked along thefibril axis so that the sidechain amides of Asn or Gln areallowed to form polar zippers, suggesting that these zippersmay stabilize amyloid-like conformations. We therefore moni-

tored the HBs between the Q22, N23, N27, and Q28 side-chains, which had the potential to form polar zippers in theEDK and DK mutations.

Our results (see Fig. 5) reveal that within 1 ns, the num-ber of water molecules in the channel is reduced to about5 � 2. Coupled with this, polar zippers form between theburied N23 and Q28 residues. There are on average about2.8 HBs [empty arrows in Fig. 5(b)] involved in the polarzippers of buried N23 (1.8 � 0.4) and Q28 (1.0 � 0.5) in theEDK and DK mutations (these values are the averagesover all EDK and DK simulations). When the simulationslast longer than 25 ns, the water molecules in the channelare reduced to about 1, and the HBs in the polar zippers ofN23 and Q28 are increased to about 6. It is clear that inthe presence of a large number of water molecules, polarzippers cannot be formed. Upon the expulsion of the watermolecules, polar zippers can be effectively formed andamyloid-like b-sheets are well maintained [Fig. 5(c)]. Onthe other hand, the sidechain amide groups of the N27 res-idues are outside the channel and are fully solvated. Theseamide groups cannot effectively form polar zippers, as indi-cated by only about 0.9 � 0.3 HBs. The amide groups ofQ22 in the EDK mutant are also outside the channel.These sidechains, which are sandwiched by the sidechainsof F20 and V24 [Fig. 5(a)], are only partially exposed to thesolvent. As a result, there are about 1.9 � 0.5 HBs amongthese amide groups.

Stability of the ED Mutant and Its Preferencefor Another Amyloid-like Conformation

We have shown that soluble oligomers with the pro-posed amyloid conformation [TopA, Fig. 1(c), left] can be

Fig. 3. The last frames of the TopA-EDK and TopA-DK simulations.

Fig. 4. Change of the percentage of in-register HBs during TopA-OPLS-WT (a) and TopA-OPLS-EDK (b) simulations. The last frame ofeach simulation is given.



more stable in the EDK simulations than in the WT sim-ulations, which agree with the experimental observationthat neutralization of buried charges may promote aggre-gation.31 Experiments have shown that the E22Q andD23N mutants can also promote oligomerization andaggregation.10,11 The ED double-mutation leads furtherto a ready formation of fibrils.30 Thus, we can assumethat ED amyloid-like oligomers are more stable than WToligomers. However, our simulations indicate that the EDhexamers with the TopA amyloid conformation have only�35% in-register HBs in the region 22–29 and �43% in-register HBs in the region 30–35 at the end (>50 ns), asshown in Table II (TopA-ED). Both the inside layers(I and II) and the outside layer (III) lose their in-registerHBs. The furthest inside layer (I) of the TopA-EDoligomer has many fewer HBs than that of WT in thefirst 5 ns of the simulations (55% vs. 70–75%). This isbecause the proposed amyloid conformation (TopA)requires that residues D23 and K28 form salt bridges. Inthe ED mutant, the anionic D23 in the WT is mutated toneutral N23. Therefore, the formation of salt bridges in

the channel becomes impossible in the TopA-ED simula-tion. On the other hand, the sidechains of K28 form a cat-ionic array and have significant electrostatic repulsions,especially in the low dielectric medium as the protein in-terior. This is indicated by the observed distances be-tween the Nf atoms of neighboring K28 residues. Asshown in Figure 6(a), these distances range from 9–18 A(numbers in bold), significantly longer than the corre-sponding distances (about 6 A) in the salt-bridged struc-tures of the WT oligomers. The distances between the Ca

atoms of the adjacent K28 residues [5.4–15.6 A, Fig. 6(a)]also become much longer than 4.8 A, a required distancefor backbones to form in-register HBs.

Thus, our simulation results appear to indicate that theED oligomers with the TopA structure are not more stablethan are the WToligomers. This agrees with experimentalobservations of Kirkitadze et al.62 It has been suggestedthat Ab peptides have multiple amyloid conformations.32

We therefore suspect that there can be another amyloid-like conformation that is more stable than the TopA con-formation for the ED mutant. The way to construct thisother amyloid-like conformation is to keep 22–29 as abend and 16–21 in contact with 30–35, which was sug-gested to be thermodynamically reasonable in previousstudies.15,17,63 Keeping this in mind, we constructed theTopB conformation based on the TopA conformation byflipping region N27-M35 and turning the original interiorside into the exposed side [Fig. 1(c), left ? right], which issimilar to the ‘‘F19/M35’’ model proposed in previous stud-ies.16,18 The difference is that K28 is no longer buriedinside the channel so that it can be solvated to avoid theunfavorable electrostatic repulsions. Our simulations indi-cate that the ED mutant with the TopB conformation hasamyloid-like oligomers that are as stable as the EDK mu-tant with the TopA conformation (Table II, TopB-ED). TheK28 sidechains become solvated so that the repulsionsbetween the neighboring Lys residues are screened. Thecharged groups of the K28 residues are now moderatelyseparated (5.4–9.7 A) and the distances between adjacentCa atoms (4.4–4.9 A) are just right for the maintenance ofthe backbones in the sheet structures [Fig. 6(b,c)].

An important feature of b-sheets is their twist, definedas y, which is the dihedral angle between neighboringb-strands [Fig. 7(a)]. Generally, b-sheets in proteins areright-hand twisted (�308 < y < 08).56 The TopA simulationresults (an average y of �108 � 48 for TopA-EDK) in thisstudy as well as in previous reports15,17,18 are in accord withthis feature [Fig. 7(b), left]. However, b-sheets in the TopBmodel are very flat with an average of y as 08 [Fig. 7(b),right]. Chothia et al. found that in b-helical structures, b-sheets do not prefer to twist if they are packed well witheach other.64 On the packing sites between regions 17–24and 28–35, the arrangements of buried sidechains ac-tually are distinct in the TopA and TopB conformations[Fig. 1(c), dark green lines for interfaces]. The two sheetsof the TopB conformation seem to pack well, with thebulky sidechains of one sheet inserting into the bucklescomposed of sidechains of the other sheet with a bulky-small-bulky pattern [Figs. 1(c) and 7(b), right]. On the

Fig. 5. (a) The top views of the DK/EDK mutant with the TopA model.(b, c) The structural change in a TopA-EDK simulation over time. Amidesidechains of N23 and Q28 are shown. Empty arrows are the HBsamong sidechain amide groups. [Color figure can be viewed in the onlineissue, which is available at www.interscience.wiley.com.]



other hand, this tight packing is absent in the TopA con-formation [Figs. 1(c) and 7(b), left]. To quantify the shapecomplementarities of the different conformations, their Scscores were computed (see Materials and Method).57 TheSc score ranges from 0 to 1 with 1 representing the perfectmatch between two surfaces. The Sc for the TopA confor-mation is 0.70 � 0.05, comparable to that of a typical flat-tened b-helical structure (SulfD: PDB 1vh4) with a normalright-hand twist; while the Sc for TopB is 0.78 � 0.04,closer to that of the ‘‘amyloid-like’’ crystal structures con-taining similarly tightly packed sheets without any twistas reported recently by Nelson et al. [Fig. 7(b), bottom].44

The TopB conformation indeed exhibits high complemen-tarity between two b-sheets, probably leading to the flatsheets, while the b-sheets in TopA are less tightly packedand have typical twists.Ma and Nussinov have proposed TopA as the Ab confor-

mation in fibrils at neutral conditions,17 which was con-

firmed by experiments.16 A very recent study by Tycko and

coworkers group reveled that Ab10–40 in fibrils actually

adopts a conformation similar to TopB at the acidic condi-

tions based on the solid-state NMR signals of residual con-

tacts.65 Our simulation indicated that in the ED muta-

tions, where E22 and D23 are neutralized, analogous to

the low pH conditions, TopB is indeed more stable than

TopA, in agreement with Tycko’s observations. Further-

more, Ab10–35 forms twisted fibrils at pH 7.4 while fibrils

of Ab10–35 and Ab10–40 have no apparent twist at acidic

conditions.34,65 The preference of twisted sheets in TopA

and the preference of flat sheets in TopB may be an impor-tant cause for such difference of fibril morphology.

Rationale for the Stabilities of Amyloid-likeConformations at Different Charged States

Burial of salt-bridges into protein interiors always encoun-ters two counteracting forces: unfavorable desolvation of

Fig. 6. The last frames of TopA-ED (a) and TopB-ED (b) simulations. The numbers in bold are the distances between Ns atoms of adjacent K28residues and the numbers in gray and italic are the distances between Ca atoms of adjacent K28 residues. (c) The changes in the distances betweenNs atoms of adjacent K28 residues in layers I–III of TopA-WT, TopA-ED, and TopB-ED simulations.

Fig. 7. (a) The twist of sheets. (b) Top views of the last frames ofTopB-ED and TopA-EDK simulations. Empty and filled arrows point tobulky and small interior sidechains, respectively.



charges and favorable enhancement of electrostaticinteractions due to lowered dielectrics. Previous studiessuggested that the net effect is a preference against bur-ial.66–68 This implies that in order to have chargedgroups inside, extra HB interactions should be presentin the proteins.66 A PDB survey found that buriedcharges always form 2–3 HBs with neutral polargroups.69 The study by Baumketner et al. found that thecarboxylic group of D23 can form on average 3.0 HBswith the backbone amides to stabilize a b-bend structureof Ab21–30 monomers.70 However, the amyloid-like con-formations in this study rarely provide such stabiliza-tion of buried D23/K28 sidechains. These chargedgroups are surrounded by buried hydrophobic side-chains and backbones of the 22–29 region [Fig. 1(c),left]. The only polar groups available to form HB with

buried D23/K28 sidechains are backbone amides in the22–29 region. The 22–29 region of amyloid-like oligom-ers in this study adopt a b-bend motif, which is quitesimilar to the proposed general feature for b-helical con-formations by recent simulations and PDB survey interms of backbone dihedral angles (see Fig. 8).63 Inthese structures, most backbone amides are involved ininter-molecular HBs63 and may not be available for bur-ied charges. Actually, when WT hexamers are in theamyloid-like structures (i.e. with 50 ns of simulations),each D23/K28 pair of nonterminal peptides (B–E) hasonly on average 1.5 � 1.3 HBs with all backbone amides,or �0.8 HBs per charge group. Probably for this reason,the water has to enter into the channel and hydrate theinterior salt-bridges to involve extra HBs, as observed inour study and the previous reports.16,18

Fig. 8. Ramachandran maps of residues in the region D23-K28. The maps are drawn based on the resi-dues in nonterminal peptides (B–E) over all WT simulations. The notes in italic indicate the backbone confor-mations of the same residues in the reported bend structure.



The extra HBs between the water molecules and theburied D23/K28 in the channel are somewhat differentfrom those between the polar groups in proteins and bur-ied ions. First, water loses its entropy when it hydrates inthe channel; second, water in the channel will contactthe hydrophobic residues, which surround the channel[Fig. 1(c), left]. These two factors reduce the stability ofthe internal hydration. Although about two water mole-cules are found for each buried D23/K28 salt-bridge in theTopA-WT simulation, this hydration is still not enough tostabilize the buried salt-bridge. Thus, further hydration isneeded to stabilize the buried D23/K28 salt-bridge. Thismay explain the observation that the number of hydrationwater molecules of nonterminal D23/K28 increase by morethan 50% (�11 ? 16–19, Table II), which is not dependenton the methods for the treatment of electrostatics, nordoes it rely on the force fields and solvent models. The fur-ther hydration of the D23/K28 sidechains does not occurby putting more water molecules into the channel becauseof the limited size of the channel (a diameter of about6–7 A). If more water molecules were added into the chan-nel, the channel would have to be enlarged and its solvent-accessible hydrophobic surfaces would have to becomelarger, which would result in destabilization. Therefore, theburied sidechains of D23/K28 have to turn to the solvent-exposed side in order to bind more water molecules, leadingto the gradual loss of amyloid-like structures.In spite of unfavorable solvation forces, the chains com-

posed of both intra-molecular and inter-molecular salt-

bridges18 make the core of the amyloid structures meta-stable (TopA-WT, Fig. 9). There are �65% HBs in thedeepest WT layer (I, between C and D) within 50 ns. Atthe same time, the distance between the Nf atoms of pepti-des C and D can be less than 6 A, which cannot be reachedin simulations of charged K28 without D23/K28 interac-tions [Fig. 6(c)]. It seems that if oligomers are largeenough, they may adequately shield water and reduce thedielectrics in their interior so that the unfavorable desolva-tion is largely compensated by the strengthening D23/K28interactions. We carried out a 45 ns simulation of a WTdecamer and found that the amyloid-like structure wasstill destabilized by solvation (see Fig. 10). Simulations ofeven larger oligomers will be conducted in the future totest this idea.

The unpaired charges in the TopA-ED hexamers, how-ever, strongly repel each other, causing rapid loss of thestructures (TopA-ED, Fig. 9). The number of HBs in themost inside HB layer (I) can drop to 55% within 5 ns(Table II), indicating that this destabilization is verystrong. Therefore, D23/K28 salt bridges should be criticalfor the buried charges in the amyloid, as previously sug-gested.16–18

In the EDK and DK mutants, neutral amide sidechainsof N23 and Q28 are buried in the channel in the TopAconformation. The TopA-EDK and TopA-DK simulationsindicate that the amyloid-like hexamers of the EDK andDK mutants are considerably more stable than in the WTconformation (Table II). There are two reasons for this

Fig. 9. Schemes of different conformations in this study. Hollow cylinders represent the environments where the charged/hydrophilic residues areburied (the labeled residues). 0, þ, and � denote the charge states of the sidechains of these residues.



observation. First, the desolvation barrier to the neutralamide sidechains of N23 and Q28 is not significant.Second and more importantly, these amide sidechainsform polar zippers in the channel, as shown in Figure 9.The hydrophobic environment actually promotes the for-mation of polar zippers because of the lowered dielectricsand the prevention of a solvent attack.71 Consideringthat even the terminal peptides (A and F) are very amy-loid-like, oligomers or probably protofibrils may be readyto grow by easy incorporation of additional peptides.

CONCLUSIONS

We have simulated various hexamers of the WT Aband its mutants with amyloid-like conformations in anaqueous environment. By doing so, we probed the stabil-ity of amyloid-like structures in the oligomer states,which are suspected to be involved in the formation ofprotofibrils.21,26 Our simulations indicate that at suchearly stages, solvation may significantly influence solubleamyloid-like hexamers of WT Ab and its mutants.The hydration of buried charged sidechains of D23/K28

can destabilize WT amyloid-like oligomers in regionsE22-G29 and L16-A21. Therefore, the burial of thesecharged sidechains may exhibit a barrier to the formationof amyoild-like structures at very initial stages of aggre-gation, in agreement with previous experiments.31 Inaddition, built from soluble oligomers, protofibrils maypossess in their early stages similar amyloid conforma-tions. They are partly destabilized in the 16–21 and 22–29 regions. This may explain the experimental observa-tion that well-formed fibrils have more stable amyloidconformations in their 24–27 and 17–21 regions than pro-tofibrils have28 and that fibrils contain more stable in-register HBs than do protofibrils.72

In the TopA-EDK and TopA-DK simulations, amyloid-like oligomers were stabilized by bypassing unfavorableburial of charges and, probably more importantly, byforming polar zippers from buried N23 and Q28, whichare promoted by the hydrophobic environment.61 Theincrease in the stability of the EDK/DK amyloid-likeoligomers accords with experimental suggestion that

neutralization of buried charges helps to pass the kineticbarrier.31

We found that the ED mutant may not prefer the amy-loid-like conformation reported for the WT fibril15 be-cause of electrostatic repulsion in the K28 charge array.Instead, we found another amyloid-like conformationthat is more suitable for the ED mutant. This conforma-tion flips region 27–35 and exposes the K28 sidechainsthat are buried in the proposed amyloid conformation.Thus, the repulsion of the K28 array is largely reducedby solvent screening. The amyloid-like conformation inthe ED mutant becomes more stable than that of the pro-posed WT conformation. Furthermore, b-sheets in thisconformation appear to be quite flat while b-sheets in theproposed WT conformation are right-hand twisted. Theseobservations indicate that Ab soluble oligomers may pre-fer different amyloid-like conformations when residues23 and 28 are in different charge states. This, in turn,may result in different morphologies of the protofibrils orfibrils.33,34

REFERENCES

1. Yankner BA. Mechanisms of neuronal degeneration in Alzheimer’sdisease. Neuron 1996;16:921–932.

2. Selkoe DJ. Neuroscience–Alzheimer’s disease: genotypes, pheno-type, and treatments. Science 1997;275:630–631.

3. Sisodia SS, Koo EH, Beyreuther K, Unterbeck A, Price DL. Evi-dence that b-amyloid protein in Alzheimer’s disease is not derivedby normal processing. Science 1990;248:492–495.

4. Yankner BA, Duffy LK, Kirschner DA. Neurotrophic and neuro-toxic effects of amyloid b protein: reversal by tachykinin neuro-peptides. Science 1990;250:279–282.

5. Jarrett JT, Berger EP, Lansbury PT. The carboxy terminus of theb amyloid protein is critical for the seeding of amyloid formation:implications for the pathogenesis of Alzheimer’s disease. Bio-chemistry 1993;32:4693–4697.

6. Geula C, Wu CK, Saroff D, Lorenzo A, Yuan ML, Yankner BA.Aging renders the brain vulnerable to amyloid b-protein neuro-toxicity. Nature Med 1998;4:827–831.

7. McKee AC, Kowall NW, Schumacher JS, Beal MF. The neurotox-icity of amyloid b protein in aged primates. Amyloid: Int J ExpClin Invest 1998;5:1–9.

8. Mason JM, Kokkoni N, Stott K, Doig AJ. Design strategies foranti-amyloid agents. Curr Opin Struct Biol 2003;13:526–532.

9. Bitan G, Lomakin A, Teplow DB. Amyloid b-protein oligomeriza-tion–prenucleation interactions revealed by photo-induced cross-linking of unmodified proteins. J Biol Chem 2001;276:35176–35184.

10. Bitan G, Vollers SS, Teplow DB. Elucidation of primary structureelements controlling early amyloid b-protein oligomerization. J BiolChem 2003;278:34882–34889.

11. Walsh DM, Lomakin A, Benedek GB, Condron MM, Teplow DB.Amyloid b-protein fibrillogenesis–detection of a protofibrillar in-termediate. J Biol Chem 1997;272:22364–22372.

12. Harper JD, Wong SS, Lieber CM, Lansbury PT. Observation ofmetastable A b amyloid protofibrils by atomic force microscopy.Chem Biol 1997;4:119–125.

13. Benzinger TLS, Gregory DM, Burkoth TS, Miller-Auer H, LynnDG, Botto RE, Meredith SC. Propagating structure of Alzhei-mer’s b-amyloid10–35 is parallel b-sheet with residues in exactregister. Proc Natl Acad Sci USA 1998;95:13407–13412.

14. Antzutkin ON, BalBach JJ, Tycko R. Site-specific identificationof non-b-strand conformations in Alzheimer’s b-amyloid fibrilsby solid-state NMR. Biophys J 2003;84:3326–3335.

15. Petkova AT, Ishii Y, BalBach JJ, Antzutkin ON, Leapman RD,Delaglio F, Tycko R. A structural model for Alzheimer’s b-amy-loid fibrils based on experimental constraints from solid stateNMR. Proc Natl Acad Sci USA 2002;99:16742–16747.

Fig. 10. The last frame of the TopA-DEC-WT simulation.



16. Petkova AT, Yau WM, Tycko R. Experimental constraints on qua-ternary structure in Alzheimer’s b-amyloid fibrils. Biochemistry2006;45:498–512.

17. Ma B, Nussinov R. Stabilities and conformations of Alzheimer’sb-amyloid peptide oligomers (Ab16–22, Ab16–35 and Ab10–35):sequence effects. Proc Natl Acad Sci USA 2002;99:14126–14131.

18. Buchete NV, Tycko R, Hummer G. Molecular dynamics simula-tions of Alzheimer’s b-amyloid protofilaments. J Mol Biol 2005;353:804–821.

19. Snyder EM, Nong Y, Almeida CG, Pauls S, Moran T, Choi EY,Nairn AC, Salter MW, Lombroso PJ, Gouras GK, Greengard P.Regulation of NMDA receptor trafficking by amyloid-b. Nat Neu-rosci 2005;8:1051–1058.

20. Walsh DM, Klyubin I, Fadeeva JV, Cullen WK, Anwyl R, Wolfe MS,Rowan MJ, Selkoe DJ. Naturally secreted oligomers of amyloid bprotein potently inhibit hippocampal long-term potentiation in vivo.Nature 2002;416:535–539.

21. Walsh DM, Hartley DM, Kusumoto Y, Fezoui Y, Condron MM,Lomakin A, Benedek GB, Selkoe DJ, Teplow DB. Amyloid b-pro-tein fibrillogenesis–structure and biological activity of protofibril-lar intermediates. J Biol Chem 1999;274:25945–25952.

22. Jablonowska A, Bakun M, Kupniewska-Kozak A, Dadlez M. Alz-heimer’s disease Ab peptide fragment 10–30 forms a spectrum ofmetastable oligomers with marked preference for N to N and Cto C monomer termini proximity. J Mol Biol 2004;344:1037–1049.

23. Harper JD, Wong SS, Lieber CM, Lansbury PT. Assembly of Abamyloid protofibrils: an in vitro model for a possible early eventin Alzheimer’s disease. Biochemistry 1999;38:8972–8980.

24. Blackley HKL, Sanders GHW, Davies MC, Roberts CJ, Tendler SJB,Wilkinson MJ. In-situ atomic force microscopy study of b-amyloidfibrillization. J Mol Biol 2000;298:833–840.

25. Nybo M, Svehag SE, Nielsen EH. An ultrastructural study of amy-loid intermediates in Ab (1–42) fibrillogenesis. Scand J Immunol1999;49:219–223.

26. Kowalewski T, Holtzman DM. In situ atomic force microscopystudy of Alzheimer’s b-amyloid peptide on different substrates:new insights into mechanism of b-sheet formation. Proc NatlAcad Sci USA 1999;96:3688–3693.

27. Mastrangelo IA, Ahmed M, Sato T, Liu W, Wang C, Hough P,Smith SO. High-resolution atomic force microscopy of solubleAb42 oligomers. J Mol Biol 2006;358:106–119.

28. Williams AD, Sega M, Chen M, Kheterpal I, Geva M, Berthelier V,Kaleta DT, Cook KD, Wetzel R. Structural properties of Ab proto-fibrils stabilized by a small molecule. Proc Natl Acad Sci USA2005;102:7115–7120.

29. Han W, Wu YD. A strand-loop-strand structure is a possible in-termediate in fibril elongation: long time simulations of amylold-b peptide (10–35). J Am Chem Soc 2005;127:15408–15416.

30. van Nostrand WE, Melchor JP, Cho HS, Greenberg SM, Rebeck GW.Pathogenic effects of D23N Iowa mutant amyloid b-protein. J BiolChem 2001;276:32860–32866.

31. Sciarretta KL, Gordon DJ, Petkova AT, Tycko R, Meredith SC.Ab40-Lactam(D23/K28) models a conformation highly favorablefor nucleation of amyloid. Biochemistry 2005;44:6003–6014.

32. Petkova AT, Leapman RD, Guo ZH, Yau WM, Mattson MP, Tycko R.Self-propagating, molecular-level polymorphism in Alzheimer’sb-amyloid fibrils. Science 2005;307:262–265.

33. Goldsbury C, Fery P, Olivieri V, Aebi U, Muller SA. Multiple as-sembly pathways underlie amyloid-b fibril polymorphisms. J MolBiol 2005;352:282–298.

34. Antzutkin ON, Leapman RD, Balbach JJ, Tycko R. Supramolecu-lar structural constraints on Alzheimer’s b-amyloid fibrils fromelectron microscopy and solid-state nuclear magnetic resonance.Biochemistry 2002;41:15436–15450.

35. Klimov DK, Thirumalai D. Dissecting the assembly of Ab(16–22)amyloid peptides into antiparallel b-sheets. Structure 2003;11:295–307.

36. Klimov DK, Straub JE, Thirumalai D. Aqueous urea solutiondestabilizes Ab(16–22) oligomers. Proc Natl Acad Sci USA 2004;101:14760–14765.

37. Hwang W, Zhang S, Kamm RD, Karplus M. Kinetic control ofdimer structure formation in amyloid fibrillogenesis. Proc NatlAcad Sci USA 2004;101:12916–12921.

38. Mousseau N, Derreumaux P. Exploring the early steps of amyloidpeptide aggregation by computers. Acc Chem Res 2005;38:885–891.

39. Lakdawala AS, Morgan DM, Liotta DC, Lynn DG, SnyderJP. Dynamics and fluidity of amyloid fibrils: a model of fibrousprotein aggregates. J Am Chem Soc 2002;124:15150–15151.

40. Urbanc B, Cruz L, Yun S, Buldyrev SV, Bitan G, Teplow DB,Stanley HE. In silico study of amyloid b-protein folding andoligomerization. Proc Natl Acad Sci USA 2004;50:17345–17350.

41. Borreguero JM, Urbanc B, Lazo ND, Buldyrev SV, Teplow DB,Stanley HE. Folding events in the 21–30 region of amyloid-b-pro-tein (Ab) studied in silico. Proc Natl Acad Sci USA 2005;102:6015–6020.

42. Cruz L, Urbanc B, Borreguero JM, Lazo ND, Teplow DB,Stanley HE. Solvent and mutation effects on the nucleation ofamyloid b-protein folding. Proc Natl Acad Sci USA 2005;102:18258–18263.

43. Jang S, Shin S. Amyloid b-peptide oligomerization in silico: dimerand trimer. J Phys Chem B 2006;110:1955–1958.

44. Nelson R, Sawaya MR, Balbirnie M, Madsen AO, Riekel C,Grothe R, Eisenberg D. Structure of the cross-b spine of amyloid-like fibrils. Nature 2005;435:773–774.

45. Berendsen HJC, van der Spoel D, van Drunen R. GROMACS–amessage-passing parallel molecular-dynamics implementation.Comput Phys Commun 1995;91:43–56.

46. van Gunsteren WF, Billeter SR, Eising AA, Hunenberger PH,Kruger P, Mark AE, Scott WRP. Biomolecular simulation: theGROMOS96 manual and user guide. Zurich: Hchschulverlag AGan der ETH; 1996.

47. Smith PE, van Gunsteren WF. Consistent dielectric-propertiesof the simple point-charge and extended simple point-chargewater models at 277 and 300 K. J Chem Phys 1994;100:3169–3174.

48. Hess B, Bekker H, Berendsen HJC, Fraaije JGEM. LINCS: a lin-ear constraint solver for molecular simulations. J Comput Chem1997;18:1463–1472.

49. Miyamoto S, Kollman PA. SETTLE–an analytical version of theshake and rattle algorithm for rigid water models. J Comput Chem1992;13:952–962.

50. Feenstra KA, Hess B, Berendsen HJC. Improving efficiency oflarge time-scale molecular dynamics simulations of hydrogen-rich systems. J Comput Chem 1999;20:786–798.

51. Tironi IG, Sperb R, Smith PE, van Gunsteren WF. A generalizedreaction field method for molecular-dynamics simulations. J ChemPhys 1995;102:5451–5459.

52. Darden T, York D, Pedersen L. Particle mesh Ewald–an N�Log(N)method for Ewald sums in large system. J Chem Phys 1993;98:10089–10092.

53. Beredsen HJC, Postma JPM, van Gunsteren WF, Di Nola A,Haak JR. Molecular dynamics with coupling to an external bath.J Chem Phys 1984;81:3684–3690.

54. Jorgensen WL, McDonald NL. Development of an all-atomforce field for heterocycles. Properties of liquid pyrrole,furan, diazoles, and oxazoles. J Phys Chem B 1998;102:8049–8050.

55. Jorgensen WL, Chandrasekhar J, Madura JD, Impey RW, Klein ML.Comparion of simple potential functions for simulating liquidwater. J Chem Phys 1983;79:926–935.

56. Chothia C. Conformation of twisted b-pleated sheets in proteins.J Mol Biol 1973;75:295–302.

57. Lawrence MC, Colman PM. Shape complementarity at protein/protein interfaces. J Mol Biol 1993;234:946–950.

58. Collaborative Computational Project, Number 4. The CCP4 suite:programs for protein crystallography. Acta Crystallogr Sect D1994;50:760–763.

59. Gursky O, Aleshkov S. Temperature-dependent b-sheet formationin b-amyloid Ab(1–40) peptide in water: uncoupling b-structurefolding from aggregation. Biochim Biophys Acta 2000;1476:93–102.

60. Liu R, McAllister C, Lyubchenko Y, Sierks MR. Residues 17–20and 30–35 of b-amyloid play critical roles in aggregation. J Neuro-sci Res 2004;75:162–171.

61. Perutz MF. Glutamine repeats and inherited neurodegenerativediseases: molecular aspects. Curr Opin Struct Biol 1996;6:848–858.

62. Kirkitadze MD, Condron MM, Teplow DB. Identification andcharacterization of key kinetic intermediate in amyloid b-proteinfibrillogenesis. J Mol Biol 2001;312:1103–1119.



63. Zanuy D, Gunasekaran K, Lesk AM, Nussinov R. Computationalstudy of the fibril organization of polyglutamine repeats reveals acommon motif identified in b-helices. J Mol Biol 2006;358:330–345.

64. Chothia C, Hubard T, Brenner S, Barns H, Murzin A. Proteinfolds in the all-b and all-a classes. Annu Rev Biophys BiomolStruct 1997;26:597–627.

65. Paravatsu AK, Petkova AT, Tycko R. Polymorphic fibril forma-tion by residue 10–40 of Alzheimer’s b-amyloid peptide. BiophysJ 2006;90:4618–4629.

66. Honig BH, Hubbell WL. Stability of ‘‘salt bridges’’ in membraneproteins. Proc Natl Acad Sci USA 1984;81:5412–5416.

67. Wimley WC, Gawrisch K, Creamer TP, White SH. Direct mea-surement of salt-bridge solvation energies using a peptide modelsystem: implications for protein stability. Proc Natl Acad SciUSA 1996;93:2985–2990.

68. Hendsch ZS, Tidor B. Do salt bridges stabilize proteins–a continuum electrostatic analysis. Protein Sci 1994;3:211–226.

69. Rashin AA, Honig B. On the environment of ionizable groups inglobular proteins. J Mol Biol 1984;173:515–521.

70. Baumketner A, Bernstein SL, Wyttenbach T, Lazo ND, TeplowDB, Bowers MT, Shea JE. Structure of the 21–30 fragment ofamyloid b-protein. Protein Sci 2006;15:1239–1247.

71. Fernandez A, Scheraga HA. Insufficiently dehydrated hydrogenbonds as determinants of protein interactions. Proc Natl AcadSci USA 2003;100:113–118.

72. Kheterpal I, Lashuel HA, Hartley DM, Walz T, Lansbury PT,Wetzel R. Ab protofibrils possess a stable core structure re-sistant to hydrogen exchange. Biochemistry 2003;42:14092–14098.



molecular dynamics studies of hexamers of amyloid-β peptide (16–35) and its mutants: influence of...

Documents