Structure and Orientation of Peptide Inhibitors Bound to Beta-amyloid ...

doi:10.1016/j.jmb.2005.09.055

J. Mol. Biol. (2005) 354, 760–776

Structure and Orientation of Peptide Inhibitors Bound to Beta-amyloid Fibrils Zhongjing Chen1†, Gerd Krause1 and Bernd Reif1,2* 1

Forschungsinstitut fu¨r Molekulare Pharmakologie (FMP), Robert-Ro¨ssle-Str. 10 D-13125 Berlin, Germany 2 Charite´ Universita¨tsmedizin D-10115 Berlin, Germany

Polymerization of the soluble b-amyloid peptide into highly ordered fibrils is hypothesized to be a causative event in the development of Alzheimer’s disease. Understanding the interactions of Ab with inhibitors on an atomic level is fundamental for the development of diagnostics and therapeutic approaches, and can provide, in addition, important indirect information of the amyloid fibril structure. We have shown recently that trRDCs can be measured in solution state NMR for peptide ligands binding weakly to amyloid fibrils. We present here the structures for two inhibitor peptides, LPFFD and DPFFL, and their structural models bound to fibrillar Ab14-23 and Ab1-40 based on transferred nuclear Overhauser effect (trNOE) and transferred residual dipolar coupling (trRDC) data. In a first step, the inhibitor peptide structure is calculated on the basis of trNOE data; the trRDC data are then validated on the basis of the trNOE-derived structure using the program PALES. The orientation of the peptide inhibitors with respect to Ab fibrils is obtained from trRDC data, assuming that Ab fibrils orient such that the fibril axis is aligned in parallel with the magnetic field. The trRDC-derived alignment tensor of the peptide ligand is then used as a restraint for molecular dynamics docking studies. We find that the structure with the lowest rmsd value is in agreement with a model in which the inhibitor peptide binds to the long side of an amyloid fibril. Especially, we detect interactions involving the hydrophobic core, residues K16 and E22/D23 of the Ab sequence. Structural differences are observed for binding of the inhibitor peptide to Ab14-23 and Ab1-40 fibrils, respectively, indicating different fibril structure. We expect this approach to be useful in the rational design of amyloid ligands with improved binding characteristics. q 2005 Elsevier Ltd. All rights reserved.

*Corresponding author

Keywords: Alzheimer’s disease; amyloid fibrils; peptide inhibitors; residual dipolar couplings; trRDCs

Introduction Alzheimer’s disease (AD) is the most common form of dementia among people above the age of 65.1 Pathologically, AD is characterized by the accumulation of amyloid plaques within the † Present address: Z. Chen, University of WisconsinMadison, Department of Chemistry, 1101 University Avenue, Madison, WI 53706, USA. Abbreviations used: AD, Alzheimer’s disease; Ab, b-amyloid peptide; EM, electron microscopy; SVD, singular value decomposition; TMS, trimethylsilane; trNOE, transferred nuclear Overhauser effect; trRDC, transferred residual dipolar coupling. E-mail address of the corresponding author: [email protected]

extracellular space of brain regions known to be important for intellectual functions.2 The major component of amyloid plaques is a 39–42 amino acid residue peptide called the b-amyloid peptide (Ab). There is increasing evidence that amyloid fibrils deposited in the brain of AD patients are linked to the pathology of the disease.3–5 Therefore, identification of small organic molecules or peptides that can prevent fibril formation, or allow detection of aggregation at an early stage of the disease is of great interest.6–8 The critical region of Ab involved in amyloid fibril formation is identified to be the hydrophobic core around residues 17–20, LVFF.9–12 Amino acid substitutions in this region by hydrophilic amino acid residues produce large changes in the peptide’s conformation and its ability to produce

0022-2836/$ - see front matter q 2005 Elsevier Ltd. All rights reserved.

761

Peptide Inhibitors Bound to Beta-amyloid Fibrils

amyloid fibrils. The shortest peptide still displaying consistently high Ab binding and amyloid formation capacity has the sequence KLVFF (corresponding to Ab16-20).11 On the basis of these results, Soto et al. suggested a lead structure for the design of inhibitors that was directed against amyloid fibril formation.13 In their strategy, the central hydrophobic region of Ab, residues 17–20 (LVFF), served as a template for designing the b-sheet breaker peptide iAb5 (LPFFD). Proline residues, which are known to have a low b-sheet formation propensity,14 were included in the peptide sequence to decrease its propensity to adopt b-sheet structure. In addition, a charged residue was added to the C terminus of the peptide to increase its solubility. Addition of a 20-fold excess of this peptide inhibitor resulted in disassembly of preformed Ab1-42 fibrils and reduced their neurotoxicity in vivo.13 It is hypothesized that binding of the b-sheet breaker peptides to Ab induces a conformational destabilization of the b-sheet structure.3 Peptide inhibitors display only weak binding with respect to b-amyloid fibrils with a dissociation constant, KD, of the order of 30 mM.15 This makes them particularly well suited for solution-state NMR studies exploiting transferred nuclear Overhauser effect (trNOE)16 and transferred residual dipolar coupling (trRDC) effects.17,18 So far, it is not clear how iAb5 can specifically interact with Ab fibrils at atomic resolution. An understanding of the detailed interaction mechanism requires the elucidation of the three-dimensional structure of the fibril/inhibitor complex. Amyloid fibril structure determination has received considerable attention over the past several years. It was discovered that, despite large differences in the size, primary structure and function, many proteins and peptides form amyloid fibrils of remarkably similar morphology and properties.19 Recently, structural information on amyloid fibrils became available from solid-state NMR experiments.19,20 However, the underlying processes whereby soluble proteins or peptides polymerize to form insoluble assemblies are poorly understood. Here, we investigate interactions between peptide inhibitors and Ab fibrils that provide valuable indirect information on fibril structure and the mechanism of fibril formation. In addition to iAb5, we study a homologous peptide iAb5inv (which has the primary sequence DPFFL), which we find to have a similar inhibitory effect on Ab1-40 and Ab14-23 fibrils. Ab14-23 was characterized to fold into an antiparallel b-sheet structure,21 whereas Ab1-40 consists of parallel b-sheets.22 The fact that the structures of Ab1-40 and of Ab14-23 in the aggregated state are different allows a comparison regarding the interaction mechanism. The NMR experimental results were used to generate a model of the structure of the peptide inhibitors bound to Ab fibrils. The structural characterization of the peptide inhibitor is based on trNOE restraints. The relative orientation of the peptide with respect to amyloid fibrils is obtained from trRDC measurements. Previously,

we showed that trRDC data can be obtained for ligands weakly aligned due to interactions with b-amyloid fibrils.23 This effect is based on the spontaneous alignment of amyloid fibrils in an external magnetic field.24,25 The alignment tensor extracted from the trRDC data is used in molecular dynamics calculations to restrain the orientation of the peptide with respect to the b-amyloid fibril. We believe that our structural data will contribute to an improved understanding of the mechanism of amyloid formation, and we expect this approach to be useful in the rational design of amyloid ligands with improved binding characteristics.

Results Electron microscope studies of Ab14-23 and Ab1-40 Figure 1(a) and (d) show electron microscope (EM) images of negatively stained Ab14-23 and Ab1-40 fibrils formed at pH 4. Acidic pH values promote aggregation and increase the rate of fibril formation.10,26,27 It was the aim to avoid conditions under which Ab molecules could possibly undergo chemical exchange between a fibrillar form and an oligomeric aggregation state. This situation was observed for Ab1-40 dissolved in aqueous buffer at pH 7,12 and for the N-terminal part of the yeast prion protein Sup35.28 Chemical exchange would complicate the analysis of the trRDC data, since intermediate Ab aggregation states would have to be taken into account. Even though fibril morphology and fibril mass per length depend on pH,29 recent EM and solid-state NMR studies carried out for Ab at pH 7.4 and 3.7 yield the same in-register, parallel b-sheet organization.30 According to Tycko and co-workers, fibrils formed at pH 7.4 correspond apparently to pairs of protofilaments or higher-order bundles, whereas single protofilaments are obtained at pH 3.7. We assume, therefore, that the acidic pH value used in our studies is in accordance with the published model of the amyloid fibril structure.22 The low pH value might even have physiological relevance, since Ab generation from the amyloid precursor protein appears to involve acidic cellular compartiments like endosomes and lysosomes.31 Taking into account that fibrils form fastest in the pH range from 3.5 to 6.0,27 it seems possible that the pH optimum of the in vitro reaction is both convenient and physiologically justified. Experimentally, we find that both Ab14-23 and Ab1-40 fibrils display straight, unbranched, ribbonlike morphologies that are 10–25 nm in diameter and several micrometers in length (Figure 1(a) and (d)). No apparent difference between fibril morphologies between Ab14-23 and Ab1-40 fibrils can be observed. This is consistent with previous observations.21

762


Figure 1. Influence of peptide inhibitors on fibril morphology of Ab studied by EM. Ab alone and with peptide inhibitor iAb5 or iAb5inv was incubated in buffer at pH 4 as described in Materials and Methods. (a) Ab1-40 alone; (b) Ab1-40CiAb5; (c) Ab1-40CiAb5inv; (d) Ab14-23 alone; (e) Ab14-23CiAb5; (f) Ab14-23CiAb5inv.

EM studies on the effect of the peptide inhibitor iAb5 and iAb5inv on Ab fibril formation The effect of the peptide inhibitors iAb5 (LPFFD) or iAb5inv (DPFFL) on the assembly process of Ab

was investigated by EM using negative staining. iAb5 was found to be able to inhibit Ab1-42 fibrillogenesis, and to disassemble preformed Ab fibrils in vitro.13 Our studies confirm this inhibitory effect of iAb5 on preformed Ab 1-40 fibrils.

763


Figure 1(a) shows a control sample, in which Ab1-40 was incubated alone in buffer at pH 4, 10 mM PO3K 4 . Only mature fibrils are observed. In contrast, large amounts of non-fibrillar and amorphous aggregates with only few mature fibrils are detectable over prolonged incubation of pre-fibrillized Ab1-40 (seven days) with a 50-fold excess of iAb5 (Figure 1(b)). In this case, the EM experiment was performed seven days after addition of iAb5 to fibrillized Ab. Generally, while mature fibril formation is decreased, the occurrence of structural intermediates is increased in the presence of the peptide inhibitors. Moreover, we find that this inhibitory effect of iAb5 is not restricted to Ab1-40, but is observable also in the case of synthetic Ab14-23. Mature fibrils are the predominant structure in the control sample of Ab14-23 (Figure 1(d)). Under similar incubation conditions, addition of a tenfold excess of iAb5 to the seven day-preformed Ab14-23 fibrils dramatically changes the morphology of the aggregates as detected seven days after the coincubation by EM (Figure 1(e)). We observe a reduced amount of fibrils. These assemblies are, in general, shorter and have a length of around 500 nm, indicating that fibrils are disassembled into smaller units. Additionally, Ab14-23 co-incubated with iAb5 appears not to form well-ordered fibrils. Instead, only short rod-like structures and non-fibrillar aggregates are observed in EM images (data not shown). A similar inhibitory effect is observed for iAb5inv, as shown in Figure 1(c) and (f). Compared to iAb5, the hydrophobic and hydrophilic amino acid residues at the N terminus and at the C terminus are exchanged in iAb5inv. In Figure 1(c), the sample was prepared by addition of a 50-fold excess of iAb5inv to Ab1-40 fibrils (seven day incubation), and detected seven days later by EM. In the EM image,

only short fibrils and pseudospherical structures are visible.32 Similarly, iAb5inv suppresses the assembly of Ab14-23 peptides into mature fibrils at a molar ratio of 10:1 (Figure 1(f)). The absence of visible elongated mature amyloid fibrils is direct evidence for the fact that the peptide inhibitors iAb5 and iAb5inv can dissolve Ab peptide aggregates. The inhibitory effect of b-sheet breaker peptides has been studied extensively by Soto and coworkers.3,13,33–36 Our aim is to contribute structural information by comparing the EM results with our NMR data, and to obtain a structural representation of the interaction mechanism from the NMR data. NMR studies NMR experiments were performed on iAb5inv in the absence and in the presence of Ab14-23/Ab1-40, respectively. The 1H 1D spectra of iAb5inv in the absence and presence of Ab14-23 fibrils are represented in Figure 2. The expanded amide and aliphatic regions of the spectra show that the peak intensity in the mixture sample is decreased by 45% compared to the peak intensity in the pure iAb5inv sample. This indicates that approximately half of all iAb5inv molecules are associated with Ab14-23 fibrils and cannot be detected. In a typical NMR binding experiment, chemical shift changes are interpreted as a function of ligand concentration. According to the Langmuir equation, the fraction of bound ligand fb is given as: fb Z

½Lfree Kd C ½Lfree

A loss of 45% of the proton yields a concentration of free ligand [Lfree] of approximately 3.25 mM, and correspondingly a dissociation constant Kd of ca

Figure 2. The 750 MHz 1H NMR spectra of 5 mM iAb5 in the absence and in the presence of Ab14-23 fibrils in 10 mM phosphate buffer at pH 4.0, 308C. The concentration of Ab14-23 was 0.5 mM.

764 3.36 mM. This value is in good agreement with Kd values reported by Kiessling and co-workers, who obtained a relative binding affinity of 1.4 mM for the peptide KLVFF in surface plasmon resonance studies. 15 The dissociation constant can be decreased to as low as 30 mM if a lysine anchor repeat sequence is appended to the sequence KLVFF. The peptide DPFFL gave very weak responses in the Biacore studies reported by Cairo et al.15 This is in agreement with our results, as we find larger values for the dissociation constant in the case of the peptide DPFFL. No resonance originating from Ab 14-23 is observed due to its large molecular mass in the fibrillized state. Care was taken to reproduce the same iAb5inv concentration in both samples (see Materials and Methods). Qualitatively, similar results were observed for iAb5inv/Ab1-40, iAb5/ Ab14-23, and iAb5/Ab1-40 systems. The identification of the spin systems for iAb5 or iAb5inv was achieved unambiguously using total correlation spectroscopy (TOCSY). The assignment of the resonances to individual amino acids was accomplished by the combined analysis of TOCSY and nuclear Overhauser effect spectroscopy (NOESY) spectra. trNOE measurements The trNOE experiments have been recorded in order to characterize the structure of iAb5 or iAb5inv in the bound state with respect to Ab fibrils. Figure 3 represents a column along the aromatic ring resonance Phe3 Hd of the NOESY spectra recorded on iAb5inv (a) without and (b)–(d) with amyloid Ab14-23 fibrils, respectively. Whereas only a few negative cross-peaks (with respect to the positive


diagonal peaks) were detected for the free peptide in solution, an extensive set of positive cross-peaks was detected when the peptide was measured after addition of Ab14-23 fibrils. Already at a mixing time of tmZ150 ms, cross-peaks between almost all 1H resonances are observed (Figure 3(d)). The observation of trNOEs confirms the binding of the peptide to Ab14-23 fibrils. In an attempt to obtain more accurate distance information from the trNOE experiment, we measured the trNOEs as a function of the mixing time. Data sets using 50 ms, 100 ms and 150 ms for 1H,1H mixing were selected for the analysis of the trNOE effects for iAb5inv and iAb5inv bound to Ab14-23 and iAb5inv bound to Ab1-40, respectively. Structure of iAb5/iAb5inv in the bound state We collected 35/40 intraresidue and 75/38 interresidue NOE contacts for iAb5/iAb5inv in the presence of Ab14-23, respectively. In the case of iAb5inv in the presence of Ab1-40, 45 intra-residue and 43 inter-residue NOE restraints were obtained. All structure calculations, minimizations and simulated annealing procedures were carried out using CNSsolve, version 1.0.37 From 200 calculated structures, 20 sets were selected for each calculation based on their energies. These structures show a high degree of similarity (Figure 4; Table 1). The sequences of two peptide inhibitors iAb5 and iAb5inv share a common core sequence (PFF), but possess exchanged N-terminal (L for D) and C-terminal (D for L) residues. The superposition of NMR structures of the two peptides iAb5 and iAb5inv bound to Ab14-23 fibrils show different backbone structures in the bound state, but remarkably similar orientations of side-chains

Figure 3. The 1D-traces along Phe3 Hd in the NOESY spectra recorded for iAb5 (a) without and with Ab fibrils ((b)–(d)) using different mixing times. The reference spectrum without Ab fibrils was recorded at 600 MHz, whereas all other spectra were recorded at 750 MHz magnetic field strength.

765


Figures 4 (a) and (b) (legend next page)

with identical properties (Figure 4(d)). Superposition of the two structures, iAb5 and iAb5inv, reveals that the side-chains of L1 and D5 of iAb5 are located at the positions of L5 and D1 in iAb5inv, respectively. This might indicate an identical spatial binding pattern for the different peptides at probably the same binding site of Ab14-23. iAb5inv binds to fibrils formed from Ab14-23 and Ab1-40, respectively, adopting different structures displaying different side-chain orientations for FFL, as shown in the superposition of iAb5inv bound to Ab14-23 and Ab1-40 fibrils (Figure 4(e)). This indicates a different geometry with respect to binding to Ab14-23 and Ab1-40 fibrils.

Measurement of residual dipolar couplings in peptide inhibitors bound to Ab fibrils Determination of the ligand geometry in the bound state requires long-range contacts between ligand and protein. Residual dipolar couplings (RDCs) can provide useful information with distance-independent angular restraints. Measurement of RDCs for the peptide inhibitor, which binds transiently to magnetically aligned fibrils, requires a net orientation of the peptide. We have shown that alignment of iAb5inv can be induced by transient binding of the peptide inhibitor to Ab fibrils.23 Ab fibrils orient spontaneously in the

766


Figure 4. Stereodiagrams of the 20 sets of lowest energy structures of the peptide inhibitor in the bound state. (a) iAb5 bound to Ab14-23; (b) iAb5inv bound to Ab14-23; (c) iAb5inv bound to Ab1-40; (d) superposition of the two average structures from (a) and (b); (e) superposition of the two average structures from (b) and (c).

magnetic field with the hydrogen bond direction (fibril axis) parallel with the magnetic field direction.24,25 The experiments were carried out as described.23 Briefly, iAb5inv was added to pre-

formed Ab14-23 fibrils. A molar ratio of iAb5inv to Ab14-23Z10:1 or iAb5inv to Ab 1-40Z50:1 was employed. The large excess of the peptide inhibitor is in agreement with the reported excess of inhibitor

Table 1. Structure calculation statistics 19

iAb5inva

iAb5invb

110 35 75 0.063

78 40 38 0.001

88 45 43 0.023

0.0097 0.821 0.227

0.0007 0.606 0.077

0.0032 0.665 0.195

F-iAb5

Number of distance restraints Intraresidue Interresidue ˚) RMSD from experimental distance restraints (A RMSD from idealized covalent geometry ˚) Bond lengths (A Bond angles (deg.) Impropers (deg.) a b

iAb5 was bound to Ab14–23. iAb5 was bound to Ab1–40.

767


required to prevent neurotoxicity to cultured neuronal cells.13 Concentrations of 0.5 mM and 0.1 mM were used for Ab14-23 and Ab1-40, respectively. Larger amounts of fibrils in the NMR tube induced significant line–broadening. Each trRDC or trNOE measurement requires ca 4 h. The RDC values and trNOEs that we observe are reproducibly the same within a period of one week. Approximately one week after incubation of fibrillar Ab14-23 or Ab1-40 together with iAb5inv, neither trNOEs nor trRDCs are observed. This is due to disaggregation of Ab14-23 or Ab1-40 fibrils by iAb5inv, which is also reflected in the loss of fibrillar material as observed by EM. All spectra were recorded on a Bruker 750 MHz solution-state NMR spectrometer, at 30 8C. One-bond 1H– 13C/ 1H–15N couplings were measured for iAb5inv in the free and bound state from 1H–13C/1H–15N HSQC spectra acquired without heteronuclear decoupling in the proton dimension. trRDCs were determined as the difference in couplings measured for the isotropic state and the bound state. All experimental trRDCs are listed in Table 2. A total of three backbone 15N–1H and five backbone 13Ca–1Ha trRDCs were measured for iAb5inv (15N, 10% 13C-labeling) in the presence of Ab14-23. For iAb5inv (13C natural abundance) in the presence of Ab1-40, a total of ten 13C–1H trRDCs were obtained. In order to estimate the error of the extracted RDC values, experiments were carried out twice using the same composition of the sample. Values reported in Table 2 correspond to the average trRDC values. The error for Ca–Ha trRDCs can be estimated to be of the order of G0.5 Hz, the error for the N–HN trRDCs is found to be of the order of G0.2 Hz. The same experiments were performed on iAb5. However, as mentioned previously, due to the different affinities of iAb5 and iAb5inv with respect Table 2. Experimental residual dipolar couplings measured for iAb5inv bound to Ab14-23/Ab1-40

Asp1 Ca–Ha Pro2 Ca–Ha Pro2 Cg–Hg Pro2 Cd–Hd Phe3 N–HN Phe3 Ca–Ha Phe3 Ch–Hh Phe4 N–HN Phe4 Ca–Ha Phe4 Cd–Hd Phe4 Ch–Hh Leu5 N–HN Leu5 Ca–Ha Leu5 Cg–Hg

iAb5inv bound to Ab14-23a

iAb5inv bound to Ab1-40b

Observed RDCs (Hz)

Observed RDCs (Hz)

K0.7 K1.1

K1.425 C1.15,C2.6 C2.175,C1.875

C0.14 C3.5 K0.83 C2.65 C3.2 K1.875 K2.92 C0.92 C0.9 K4.725

a N–H and C–H trRDCs for iAb5inv interacting with Ab14-23 employing a molar ratio of 10:1 for iAb5inv : Ab14-23. b C–H trRDCs for iAb5inv interacting with Ab1-40 employing a molar ratio of 100:1 for iAb5inv : Ab1-40.

to Ab fibrils, no significant trRDCs for iAb5 could be observed. Determination of the order tensor for iAb5inv bound to Ab fibrils Using the NOE–derived structure for iAb5inv and the observed trRDCs for iAb5inv in the presence of Ab fibrils (Table 2), the five elements of the order matrix can be determined using the singular value decomposition (SVD) approach described by Prestegard and co–workers.38 The calculation was performed with the aid of the software package PALES developed by Zweckstetter & Bax;39 the order parameters are listed in Table 3. The set of allowed solutions from the SVD analysis is plotted as a distribution of points on a Sauson–Flamsteed plot, as shown in Figure 5. The blue, green, and red spots depict the direction of the axes (z, y, and x, respectively) of the principle alignment frame relative to the starting structure coordinate frame. The top and bottom tips of the map represent CZ and KZ in the starting structure coordinate frame, while CX is in the very center. In Figure 5, we find two clusters of solutions that are symmetric with respect to inversion on the origin. The alignment tensor is extremely asymmetric and the directions of all three principal axes of the alignment tensor are well defined. The asymmetry is defined quantitatively by an asymmetry parameter h, which can be written in terms of order parameters Sij ðhZ ðSyy KSxx Þ=Szz Þ. We find h equal to 0.67 and 0.49 for iAb5inv bound to Ab14–23 and Ab1–40, respectively. The alignment tensor of the bound ligand is used to restrain the absolute orientation of the ligand in the ligand/fibril complex. As suggested by Prestegard and co-workers,40 in the case of a rigid ligand–protein complex, the directions and levels of the orienting force should appear the same from the point of view of the ligand and the protein. Therefore, one would speculate that the alignment tensor of a ligand and a protein should be the same in the case of a rigid complex. Considering that Ab fibrils align in the magnetic field,25 the direction of highest order (Szz) depicted in Figure 5 should coincide with the fibril axis, parallel with the external magnetic field. Obviously, iAb5inv adopts a different orientation when bound to Ab14-23 and Ab1-40, respectively. These relative different orientations were used subsequently as starting points for the docking studies in the molecular dynamics simulations. Validation of trRDCs for iAb5inv by back-calculation of RDCs In order to validate the experimental trRDCs obtained for iAb5inv bound to Ab fibrils, the measured RDCs were back-calculated on the basis of the NOE-derived structures and the abovedetermined alignment tensors. Figure 6 shows a correlation between the trRDCs measured for

768


Table 3. Calculated order parameters for iAb5inv bound to Ab14-23/Ab1-40, respectively

Number of successful steps Number of used RDCs Order matrix (Saupe average) Principal order matrix Da (Z1/2 Szz) Dr (Z1/3(SxxKSyy) RMS Correlation rate

iAb5inv bound to Ab14-23

iAb5inv bound to Ab1-40

475 8 K1.0949!10K4 3.3207!10K4 K2.0484!10K5 K6.3617!10K5 1.7826!10K4 4.7707!10K5 2.4279!10K4 K2.9049!10K4 K1.305149!10K5 K7.940946!10K5 0.868 0.972

79 10 1.3645!10K5 1.8146!10K4 K9.8683!10K5 K8.9565!10K5 3.9474!10K5 K4.2724!10K5 K1.4087!10K4 1.8360!10K4 8.920302!10K5 2.664430!10K5 1.152 0.961

Figure 5. Orientations of the order tensor principal axis systems in the molecular frame as determined via SVD. (a) and (b) show the results for iAb5inv in the presence of Ab14-23 and Ab1-40. The Sauson–Flamsteed projection maps the surface of a unit sphere into a plane by converting latitude (f) and longitude (l) to Cartesian coordinates (x, y) via xZ l cosf, yZf. The horizontal lines of latitude run from K908 to 908 in 108 increments. Vertical curved lines of longitude run from K1808 to 1808 in 208 increments. Each point in these plots represents the location, in the molecular frame, of the tip of the x, y, or z unit vectors of the order tensor principal axis system.

Figure 6. Correlation plot of experimentally measured RDCs versus predicted RDCs from the NOE-derived structures. (a) iAb5inv bound to Ab14-23;. (b) iAb5inv bound to Ab1-40.

iAb5inv in the presence of Ab fibrils and the RDCs calculated from the NOE-derived structures. The slope, which is close to 1, and the modest scatter (the correlation coefficient of 0.972 and 0.962 for iAb5inv bound to Ab14-23 and Ab1-40, respectively) suggest that the trRDCs represent well the structure in the bound state, which confirms that the measured trRDCs are not due to experimental errors. In principle, correlation plots of structurally

predicted RDCs versus measured RDCs would be expected to have a slope of 1 (continuous lines in Figure 6(a) and (b)). The slopes fitted here are, however, less than 1 (broken lines in Figure 6(a) and (b)). This attenuation has been attributed to structural uncertainty that diminishes the apparent degree of alignment. Deviations of the measured RDCs from those calculated from a structure can arise from either static or dynamic differences.


Especially, the coordinates used were not refined with other restraints except for trNOEs. Dynamic structural averaging over time-scales ranging from pisoseconds to several milliseconds can affect the value of the experimental trRDCs as well.41

Discussion The peptide inhibitors iAb5 and iAb5inv were docked to Ab fibrils as described in Materials and Methods. As a suitable template for fibrillar Ab, we used the structural model described by Tjernberg et al.,21 in the case of Ab14-23. For Ab1-40 fibrils, we used the solid-state NMR structure determined by Petkova et al.22 Ab1-40 fibrils are characterized by parallel in-register b-strand orientations, whereas Ab14-23 fibrils adopt an antiparallel b-strand orientation, which folds into a 17Ck421Kk registry (i.e. intermolecular hydrogen bonds between residues 17Ck and 21Kk of adjacent peptide chains, for integer k). Tjernberg et al. studied the registry of hydrogen bonding in antiparallel b-sheets for Ab1423 fibrils based on a combination of statistical analysis predictions reported by Wouters & Curmi42 and experimental studies of the aggregation of b-hairpin peptides constrained to favor particular registries.43 The results indicate that, in the context of designed b-hairpins, Ab14-23 can adopt both 17Ck420Kk and 17Ck421Kk registries in amyloid fibrils at pH 7.4. Experimentally, a 17Ck421Kk registry was observed in Ab16-22 fibrils at pH 7.0.44 More recently, Petkova et al. reported a pH-dependent antiparallel b-sheet registry in fibrils formed by Ab11-25.45 In their study, 17Ck420Kk and 17Ck422Kk registries were determined based on experimental solid-state NMR data for Ab11-25 at pH 7.4 and pH 2.4, respectively. Furthermore, Tycko and co-workers analyzed the registry of antiparallel b-sheets using the information theory formalism provided by Steward & Thornton,46 and found 17Ck421Kk and 17Ck420Kk as the two most likely registries for Ab16-22 and Ab11-25 fibrils at pH 7.4. Substituting E22 by Q22 and D23 by N23 in the peptide sequence to mimic the effects of low pH in the calculation, results in an increased likelihood for 17Ck421Kk and 17Ck422Kk registries. We assume, therefore, that the 17Ck421Kk registry in Ab14-23 at pH 4 suggested by Tjernberg et al.21 provides a good

769 model to describe the structure of Ab14-23 fibrils at pH 4 in the absence of solid-state NMR experimental data. Figure 7 shows the docking models of iAb5 and iAb5inv with Ab14-23, respectively. In the structure of iAb5 (Figure 7(a)), the side-chain of L1 interacts with F20(I) and the N terminus of iAb5 is interacting with E22(I), and the carboxylic acid of D5 shows an interaction with K16(IC1) of Ab14-23 fibrils. Here, I or IC1 refer to successive b-strands in the fibril structure. For iAb5inv (Figure 7(b)), D1 and L5 are interacting with K16(IC1) and F19I, respectively. Note that although the side-chain orientation of iAb5 and iAb5inv differs, the side-chains of D5 in iAb5 and D1 in iAb5inv as well as L1 in iAb5 and L5 in iAb5inv interact with the same chemical groups in Ab14-23 fibrils. The docking is in agreement with the NOE-derived peptide structures, which indicated that the side-chains in both peptides are superimposable. We obtain the lowest energies for docking, if iAb5 and iAb5inv are interacting at the same binding sites of Ab14-23. Comparing iAb5 and iAb5inv, we find that iAb5 displays more interactions to Ab14-23 than iAb5inv (one more electrostatic interaction). This might be reflected also in the trNOESY cross-relaxation data, which contain indirect information on the dissociation constant Kd of the ligand–fibril complex: we observe larger cross-peaks for iAb5 bound to Ab14-23 compared to the case when iAb5inv is added to preformed Ab14-23 fibrils. Comparing the structures of iAb5inv bound to Ab14-23 and Ab1-40 fibrils, we find different conformations for iAb5inv. The different structure can only be due to differences in the fibril structure of Ab14-23 and Ab1-40. The different interaction patterns are clearly visible in the docking models. An antiparallel organization of the b-strands along the fibril axis was characterized for the short amyloid peptide Ab14-23. The central residues L17VFFA form a hydrophobic core at a single molecular layer. This hydrophobic block is flanked by alternating patterns of positive and negative charges caused by the charged C-terminal (K16) and N-terminal (E22 and D23) residues. These systematic alternations of charges are responsible for interaction of such building blocks along and orthogonal to the fibril axis. The parallel arrangement of b-strands along the fibril axis for the fulllength amyloid peptide Ab1-40 shows a completely

Figure 7. Docking models of (a) iAb5 (yellow) and (b) iAb5inv (gray) bound to Ab14-23, respectively. For the orientation of iAb5 inv, the trRDC data were used.

770 different interaction pattern for possible ligand binding: Ab1-40 fibrils consist of two b-strands that are separated by a 1808 bend. The b-strands form two in-register parallel b-sheets that interact through mainly intramolecular hydrophobic sidechain contacts. At the surface of a single molecular layer (cross-b unit), the positively (K16) and negatively charged (D22) residues of the first b-strand (9–22) are arranged with respect to each other in a row along the fibril axis. The two charged rows are separated by a hydrophobic and an aromatic row (V18, F20). Hydrophobic side-chains of the second b-strand (30–40) form a hydrophobic face at the opposite side of the cross-b unit. As in the case of Ab14-23, we would expect similar interactions for iAb5 and iAb5inv with respect to Ab1-40. However, in the case of iAb5, we observed very broad resonance lines in the 1H spectrum. This might be due to a different number of interactions between iAb5 and Ab1-40 compared to iAb5inv/ Ab1-40, which might lead to a different dissociation constant Kd. Two general orientations of iAb5inv towards the hydrophilic phase of the Ab1-40 fibril surface are then allowed (Figure 8). The orientation in which the two negative charges of the ligand are oriented towards the row of positively charged lysine residues (K16) is energetically preferred compared to the opposite ligand orientation (rotation of iAb5inv around the fibrils axis), in which the N-terminal positive charge is interacting with the train of negative charges (E22) on the fibril surface. The negative charges of D1 and the C terminus of the ligand iAb5inv are interacting with K16(IC2) and K16(IC3) of the fibril, respectively. Furthermore, the aromatic residues F3 and F4 of the ligand are in van der Waals contact with a train of aromatic residues, F20(I), F20(IC1), and F20(IC2), located in three successive b-strands. The hydrophobic residue L5 is buried between two subsequent hydrophobic residues V18(IC2) and surrounded by aromatic residues.


Diffusion experiments were carried out in order to obtain information about a possible oligomeric arrangement of the peptide inhibitor in the absence of Ab fibrils. We used trimethylsilane (TMS) as a molecular mass reference to estimate the diffusion coefficient and, deduce the molecular mass of iAb5 at different concentrations of peptide (Figure 9). Assuming a molecular mass of 88 Da for TMS we can estimate the molecular mass of [U-13C,15N]iAb5 to be of the order of 690 Da, which corresponds quite well with the expected value for the monomer (675 Da). This value is rather independent of the concentration employed. Thus, viscosity effects on the correlation time tc can be excluded, which might affect relative cross-peak intensities in NOESY spectra. The high concentration of iAb5 with respect to Ab fibrils implies that iAb5 might form oligomers in the bound state in order to dissolve Ab aggregates. In order to investigate this possibility, a sample was prepared using a 5 mM mixture of [U-13C,15N]iAb5 and natural abundance iAb5 (molar ratio 1:2) in the presence of 0.5 mM Ab14-23 fibrils. A NOESY spectrum was recorded for this sample without heteronuclear decoupling in the direct and indirect dimensions. Figure 10(a) shows a representation for the expected cross-peak patterns. Resonances originating from unlabeled and labeled molecules are marked as filled and open circles, respectively. In the absence of intermolecular contacts, no cross-peak at the position of the broken circle would be expected, and this is indeed observed experimentally (Figure 10(b)) for most of the resonances. We find only a very weak correlation between an aromatic proton in one molecule to a Leu-CH3 in a neighboring molecule in the bound state (Figure 10(c), indicated by an arrow). We therefore speculate that iAb5 interacts as an oligomer with Ab fibrils. These iAb5$iAb5 interactions are very difficult to observe, since the observation of intermolecular NOEs requires a life-time of the oligomeric state of z5 ms.

Figure 8. The model of iAb5inv bound to Ab1-40 fibrils using the orientational restraints from trRDC data.


771

Figure 9. Diffusion ordered spectroscopy (DOSY) experiments carried out for iAb5 at different concentrations in the absence of Ab fibrils. Trimethylsilane (TMS) was used as a molecular mass standard. No concentration dependence of the magnetization decay is observed, indicating that iAb5 is monomeric in solution.

Figure 10. Expanded region of a NOESY spectrum recorded for a 1:2 mixture of (U-13C, 15N)-labeled iAb5 and iAb5 in the presence of Ab14-23 fibrils. (a) A representation of the expected NOESY spectrum. Broken circles indicate possible intermolecular iAb5$ iAb5 correlations. (b) Aliphatic region of the experimental NOESY spectrum. (c) Region displaying correlations between aromatic and aliphatic resonances. The 1D traces were taken at the indicated positions.

772


Figure 11. (a) Orientation of a C–H vector of the ligand peptide in the magnetic field B0 assuming a mosaic spread of bmosaic. qF,CH describes the angle between the fibril axis and the C–H vector. Due to fibril symmetry, the C–H vector can adopt all possible values of f on the surface of a cone around the fibril axis. (b) Effect of mosaic spread on the experimental RDCs.

In order to estimate the effect of fibril disorder in the NMR tube on the measured trRDC values, we carried out simulations that show that the extracted angular information is largely independent of the amount of fibril misalignment. Similar to the analysis of aligned samples in static solid-state NMR experiments,47,48 a descriptor “mosaic spread” can be introduced that allows us to take into account the deviation of the amyloid fibril axis from the axis of the external magnetic field (Figure 11(a)). The angle b0,C–H, which describes the orientation between an arbitrary C–H bond vector of the ligand and the magnetic field axis, can be expressed as a function of the disorder angle bmosaic between the fibril axis and the axis of the external magnetic field B0, the angle qF,CH between the fibril axis and the C–H vector in the ligand, and a polar angle f:49 cos bB0 ;CKH Z cos bmosaic cos qF;CKH Ksin bmosaic sin qF;CKH cos 4 Integration over all possible values of f yields a dependence of the experimental residual dipolar coupling on the orientation of the C–H dipole vector of the ligand to the fibril axis. As one can see from Figure 11(b), systematic deviations for the measured

trRDCs are expected only if the orientation of the respective bond vector is parallel with the fibril axis (10% error, assuming 158 mosaic spread for parallel orientation). The effect is, however, almost negligible for orientations 548!qF,CH !1278. Mosaic spread will contribute to the line-width of the measured multiplet. Experimentally, we find that the line-width is not changed significantly in the presence or in the absence of amyloid fibrils. We assume that this is due to the small RDC values that we observe, which are smaller than the natural line-widths of the nitrogen line-width. The NMR data presented here do not allow us to determine whether ligand binding occurs as well on the tips of the amyloid fibril, since the NMR experiments reflect an ensemble average and therefore would not detect possible interactions with less populated binding sites. In fact, molecular docking studies show that the experimental NMR restraints obtained are not consistent with a model in which the inhibitor binds to the tips of a growing amyloid fibril. We speculate that the interaction of peptide inhibitors with the tips of fibrils might be more relevant with respect to the inhibitory effects, as amyloid fibrils seem to grow by monomer addition.50,51 Goto and co-workers have shown that a peptide inhibitor consisting of Ab25-35 with D-amino acid residues in defined positions binds to

773


the tips of Ab25-35 fibrils and efficiently inhibits fibril elongation. The inhibitor was, however, ineffective against fibrils formed by the full-length peptide Ab1-40. Furthermore, no inhibitory effect for peptides with b-structure breaker propensities could be observed. This might be due to competitive binding of the peptide LPFFD and thioflavin T (ThT) with respect to Ab1-40. In fact, it could be demonstrated that three distinct binding sites exist for a class of benzothiazole-derived compounds in Ab fibrils.52 Competition assays and determination of the stochiometry of ligand binding should yield more information and better insight about possible ligand-binding sites in Ab.53 Ligands that bind stochiometrically, e.g. Congo red, might intercalate between b-sheets.54,55 Ligands that bind substochiometrically would presumably recognize less common structural elements as the ends of protofibrils or interfaces between protofibrils. In conclusion, we could show by NMR and restrained ligand/fibril docking studies that iAb5 binds to the hydrophobic core of Ab. Residue K16 and F20 of the Ab amyloid fibril serve as an anchor for iAb5. Different peptide inhibitors are associated with amyloid fibrils at similar binding sites. We believe that competition for hydrophobic and electrostatic interactions, and at the same time the inability to form hydrogen-bonded b-sheet structures, yields a weakening of hydrogen bonds among subsequent b-sheets of the amyloid fibril, and therefore induces dissolution of b-amyloid fibrils. Complementary experiments will have to be carried out, however, in order to clarify the question of how the suggested ligand-binding mode avoids templating for further fibril growth, or if other mechanisms are responsible for preventing fibril elongation.

Materials and Methods Synthetic peptides Ab1-40 was purchased from BioSource International Inc., USA. Other Ab peptides were synthesized manually using standard Fmoc chemistry.56 Peptides were cleaved from the resins with a solution of 90% (v/v) trifluoroacetic acid for approximately 60 min and then precipitated from the cleavage solution in diethylether. Precipitated peptides were washed several times with diethylether, centrifuged, and dried under vacuum. Peptides were purified by HPLC at room temperature, using a Polyencap column and a water/acetonitrile gradient with 0.1% trifluoroacetic acid. Peptides were lyophilized and stored. Peptides were characterized by reverse-phase HPLC and electrospray mass spectrometry with satisfactory results in all cases. Final yields of purified peptides were roughly 60% of the theoretical maximum for iAb5 and 40% for Ab14-23. Ab fibril formation As a buffer, 10 mM PO3K 4 (pH 4.0) was employed to allow for optimal formation of fibrils.26 Fibrillar peptides

were prepared as described.57 In this protocol, lyophilized Ab14-23 is solubilized first in trifluoroacetic acid. The solutions were kept at room temperature for 1 h or 2 h, until the peptide dissolved completely. After removal of the solvent with dry nitrogen gas, Ab14-23 was resolubilized in 1,1,1,3,3,3-hexafluoro-2-propanol. A small amount of the concentrated stock solution (0.37 mg/50 ml) was transferred to the aqueous buffer (600 ml) to yield the final concentration of 0.5 mM. In the case of Ab1-40, a final concentration of 0.1 mM was employed. Electron microscopy In order to prepare Ab fibrils, lyophilized peptide powder was treated as described above, and incubated in aqueous buffer for seven days prior to the EM measurements. Mixtures of Ab fibrils and inhibitors were prepared by adding a concentrated solution of the inhibitor peptide to preformed Ab fibrils. EM measurements were performed seven days after the co-incubation. Samples were placed on grids covered by a carbonstabilized formvar film. Excess fluid was withdrawn after 30 s, and the grids were negatively stained with 3% (w/v) uranyl acetate in water. Grids were allowed to dry in air, examined and photographed using a 902A electron microscope (Carl Zeiss, Oberkochen, Germany) equipped with MegaView III camera (Soft Imaging System, Mu¨nster, Germany). Images were acquired and processed by means of the analySISw Digital Micrograph program (Soft Imaging System, Mu¨nster, Germany). Nuclear magnetic resonance spectroscopy All NMR spectra were recorded on Bruker Avance 600 MHz and 750 MHz NMR spectrometers at a temperature of 30 8C. Chemical shifts at this temperature were calibrated with respect to TMS. Amino acid spin systems were identified by two-dimensional TOCSY using a mixing time of 50 ms and NOESY using mixing times of 50 ms, 100 ms, 200 ms, and 500 ms. The assignments of the b and g protons of leucine and proline residues of the peptide were further confirmed using a 1 H,13C-heteronuclear single quantum coherence (HSQC) experiment. All NMR spectra were processed and analyzed with Xwinnmr 3.5 (Bruker). In order to prepare the NMR sample, 1.9 mg of lyophilized iAb5 or iAb5inv was dissolved in 200 ml of trifluoroacetic acid for 2 h. After removal of trifluoroacetic acid by dry nitrogen gas, the peptide was re-dissolved in 30 ml of 1,1,1,3,3,3-hexafluoro-2-propanol. This solution was used as a stock solution used for NMR sample preparation. Proton-coupled 1H–15N HSQC spectra were recorded on 15N-labeled iAb5inv in the absence and in the presence of Ab14-23. Data were acquired with a tmax of 140 ms and 273 ms in F1 and F2, respectively, yielding a data matrix size of 256!4096 complex points. Eight transients per free induction decay (FID) were acquired, and quadrature in the F1 dimension was achieved using echo-antiecho coherence order selection. Data matrices were zero-filled to 2048!16,384 points to yield a digital resolution of 0.04 Hz/point in the indirect dimension. Proton-coupled 1H–13Ca HSQC spectra of iAb5inv were recorded at 13C natural abundance in the absence and in the presence of Ab1-40 fibrils by omitting the 1H scalar decoupling in the F1 dimension. A tmax of 27 ms and 80 ms was acquired in F1 and F2, respectively, yielding a data matrix size of 256!1318 complex points, resulting in

774 acquisition times of 64 transients per FID were acquired, and quadrature in the F1 dimension was achieved in the echo-antiecho manner. Data matrices were zero-filled to 16,384!32,768 points. Structure calculations Transferred-NOESY data were used to extract distance restraints for quantitative structure calculations. The cross-peak intensities were calibrated using the crosspeak intensities of the resolved phenyl-ring Hd–H3 correlation peaks assuming an internuclear distance of ˚ . Quantitative distance are obtained according to:58 2.56 A !1=6 Vref rij Z rref Vij where V refers to the intensity of the cross-peak, and r refers to the internuclear distance. NOE-derived distance restraints were used as upper limits only. All structure calculations were carried out using CNSsolve version 1.0.37 Initial embedded structures were generated from sequence structures using the sequence annealing protocol. Default values were used for all force constants and molecular parameters that were involved. High-temperature Cartesian molecular dynamics was carried out using 0.015 ps timesteps at 50,000 K (1000 steps). Cooling was then achieved via 1000 steps (using 0.015 ps time-steps). The 200 generated structures were then energetically minimized in 1000 steps using a van der Waals energy minimization. A final minimization of ten cycles of 200 steps with an NOE scaling factor of 75 kcal molK1 and a dihedral scaling factor of 400 kcal molK1 was run. Determination of peptide alignment tensor Dipolar couplings were calculated as the difference between the coupling (1JCHC1DCH or 1JNHC1DNH) in the presence of fibrils and the isotropic couplings (1JCH or 1 JNH) in the absence of fibrils. Order tensors were determined using the experimental dipolar couplings and the associated NOE-derived structures as an input for PALES.39 For the order tensor determination of iAb5inv bound to Ab14-23 and Ab1-40, respectively, a total of eight and ten RDC values were used. Only residues that are resolved in the spectrum are evaluated. Molecular modeling of iAb5/iAb5inv docking to Ab fibrils As a structural model, we use the model provided by Tjernberg et al.,21 which is characterized by an antiparallel b-strand orientation for Ab14-23, and the structural model involving parallel b-strand orientation for Ab1-40 provided by Petkova et al.22 The NMR structures of the peptide inhibitors (iAb5 and iAb5inv) were docked to the structural model of Ab14-23 and Ab1-40 using three different approaches. For automatic docking, the docking module FlexX of the Sybyl 6.91 program package from Tripos Inc. was used. For manual docking and docking by constrained molecular dynamic simulations, functional data, complementary side-chain properties and the RDC data (for iAb5inv) constraining the relative orientation of ligand with respect to amyloid fibrils were considered as additional constraints. In detail, the Cartesian coordinate system in which the iAb5inv NMR-structure is represented (in the presence of Ab fibrils) was rotated to position the z-axis in


the direction of the main component of the alignment tensor (based on RDC data using the program PALES39). The Ab14-23 model was obtained as a PDB file from Dr David Callaway. The Ab1-40 model was constructed by relying on the published backbone torsional angles.22 In the next step, the structure was rotated in order to orient the fibril axis in parallel with the z-axis, using the program VMD.59 In order to dock the ligand to the fibril, translational motions of the ligand along x, y and z, as well as rotations around the z-axis were allowed. All assembly procedures were performed in vacuo. The three assembled complex models of the peptide inhibitors and the corresponding amyloid peptide oligomers were soaked with water in a periodic boundary box. Initially, the atoms were kept fixed to relax the water during minimization. Later on, the entire system was considered. The resulting models were minimized with an AMBER 7.0 force field and simulated with a 0.5 ns molecular dynamics run at 300 K. The geometric accuracy was analyzed by using the program PROCHECK.60

Acknowledgements This research was supported by the DFG grant Re1435/2. We are grateful to Dr D. Lorenz, FMP Berlin, for performing electron microscopy experiments.

References 1. Clippingdale, A. B., Wade, J. D. & Barrow, C. J. (2001). The amyloid-b peptide and its role in Alzheimer’s disease. J. Pept. Sci. 7, 227–249. 2. Selkoe, D. (1991). The molecular pathology of Alzheimer’s disease. Neuron, 6, 487–498. 3. Adessi, C. & Soto, C. (2002). Beta-sheet breaker strategy for the treatment of Alzheimer’s disease. Drug Dev. Res. 56, 184–193. 4. Hardy, J. (1997). The Alzheimer family of diseases: many etiologies, one pathogenesis? Proc. Natl Acad. Sci. USA, 94, 2095–2097. 5. Hardy, J. A. & Selkoe, D. J. (2002). The amyloid hypothesis of Alzheimer’s disease: progress and problems on the road to therapeutics. Science, 298, 962–964. 6. Findeis, M. A. (2000). Approaches to discovery and characterization of inhibitors of amyloid beta-peptide polymerization. Biochim. Biophy. Acta, 1502, 76–84. 7. Cohen, F. E. & Kelly, J. W. (2003). Therapeutic approaches to protein-misfolding diseases. Nature, 426, 905–909. 8. Gestwicki, J. E., Crabtree, G. R. & Graef, I. A. (2004). Harnessing chaperones to generate small-molecule inhibitors of amyloid beta aggregation. Science, 306, 865–869. 9. Hilbich, C., Kisters-Woike, B., Reed, J., Masters, C. L. & Beyreuther, K. (1992). Substitutions of hydrophobic amino acids reduce the amyloidogenicity of Alzheimer’s disease bA4 peptides. J. Mol. Biol. 228, 460–473. 10. Soto, C., Castano, E., Kumar, R., Beavis, R. & Frangione, B. (1995). Fibrillogenesis of synthetic amyloid-beta peptides is dependent on their initial secondary structure. Neurosci. Letters, 200, 105–108. 11. Tjernberg, L. O., Na¨slund, J., Lindquist, F., Johansson,

775


12.

13.

14.

15. 16. 17.

18.

19. 20.

21.

22.

23.

24. 25. 26.

27.

28.

J., Karlstro¨m, A. R., Thyberg, J. et al. (1996). Arrest of b-amyloid fibril formation by a pentapeptide ligand. J. Biol. Chem. 271, 8545–8548. Narayanan, S. & Reif, B. (2005). Characterization of chemical exchange between soluble and aggregated states of beta-amyloid by solution state NMR upon variation of the salt conditions. Biochemistry, 44, 1444– 1452. Soto, C., Sigurdsson, E. M., Morelli, L., Kumar, R. A., Castan˜o, E. M. & Frangione, B. (1998). b-Sheet breaker peptides inhibit fibrillogenesis in a rat brain model of amyloidosis: implications for Alheimer’s therapy. Nature Med. 4, 822–826. Wood, S. J., Wetzel, R., Martin, J. D. & Hurle, M. R. (1995). Prolines and amyloidogenicity in fragments of the Alzheimer’s peptide b/A4. Biochemistry, 34, 724– 730. Cairo, C. W., Strzelec, A., Murphy, R. M. & Kiessling, L. L. (2002). Affinity-based inhibition of b-amyloidtoxicity. Biochemistry, 41, 8620–8629. Ni, F. (1994). Recent developments in transferred NOE methods. Prog. NMR Spectr. 26, 517–606. Koenig, B. W., Mitchell, D. C., Ko¨nig, S., Grzesiek, S., Litman, B. J. & Bax, A. (2000). Measurement of dipolar couplings in a transient peptide fragment weakly bound to oriented photo-active rhodopsin. J. Biomol. NMR, 16, 121–125. Koenig, B. W., Kontaxis, G., Mitchell, D. C., Louis, J. M., Litman, B. J. & Bax, A. (2002). Structure and orientation of a G protein fragment in the receptor bound state from residual dipolar couplings. J. Mol. Biol. 322, 441–461. Tycko, R. (2003). Insights into the amyloid folding problem from solid-state NMR. Biochemistry, 42, 3151– 3159. Jaroniec, C. P., MacPhee, C. E., Astrof, N. S., Dobson, C. M. & Griffin, R. G. (2002). Molecular conformation of a peptide fragment of transthyretin in an amyloid fibril. Proc. Natl Acad. Sci. USA, 99, 16748–16753. Tjernberg, L. O., Callaway, D. J. E., Tjernberg, A., Hahne, S., Lillieho¨o¨k, C., Terenius, L. et al. (1999). A molecular model of Alzheimer amyloid b-peptide fibril formation. J. Biol. Chem. 274, 12619–12625. Petkova, A. T., Ishii, Y., Balbach, J. J., Antzutkin, O. N., Leapman, R. D., Delaglio, F. & Tycko, R. (2002). A structural model for Alzheimer’s b-amyloid fibrils based on experimental constraints from solid state NMR. Proc. Natl Acad. Sci. USA, 99, 16742–16747. Chen, Z. & Reif, B. (2004). Measurement of residual dipolar couplings in peptidic inhibitors weakly aligned by transient binding to peptide amyloid fibrils. J. Biomol. NMR, 29, 525–530. Worcester, D. L. (1978). Structural origin of diamagnetic anisotropy in proteins. Proc. Natl Acad. Sci. USA, 75, 5475–5477. Sikorski, P., Atkins, E. & Serpell, L. (2003). Structure and texture of fibrous crystals formed by Alzheimer’s abeta(11-25) peptide fragment. Structure, 11, 915–926. Fraser, P. E., Nguyen, J. T., Surewicz, W. K. & Kirschner, D. A. (1991). pH-dependent structural transitions of Alzheimer amyloid peptides. Biophys J. 60, 1190–1201. Kirkitadze, M. D., Condron, M. M. & Teplow, D. B. (2001). Identification and characterization of key kinetic intermediates in amyloid b-protein fibrillogenesis. J. Mol. Biol. 312, 1103–1119. Narayanan, S., Bo¨sl, B., Walter, S. & Reif, B. (2003).

29.

30.

31. 32.

33.

34.

35.

36.

37.

38.

39.

40. 41.

42.

43.

Importance of low oligomeric weight species for prion propagation in the yeast prion system Sup35/ Hsp104. Proc. Natl Acad. Sci. USA, 100, 9286–9291. Wood, S. J., Maleef, B., Hartl, T. & Wetzel, R. (1995). Physical, morphological and functional differences between pH 5.8 and 7.4: aggregates of the Alzheimer’s amyloid peptide Ab. J. Mol. Biol. 256, 870–877. Antzutkin, O. N., Leapman, R. D., Balbach, J. J. & Tycko, R. (2002). Supramolecular structural constraints on Alzheimer’s beta-amyloid fibrils from electron microscopy and solid-state nuclear magnetic resonance. Biochemistry, 41, 15436–15450. Selkoe, D. J. (1994). Cell biology of the amyloid betaprotein precursor and the mechanism of Alzheimer’s disease. Annu. Rev. Cell Biol. 10, 373–403. Goldsbury, C. S., Wirtz, S., Mu¨ller, S. A., Sunderji, S., Wicki, P., Aebi, U. & Frey, P. (2000). Studies on the in vitro assembly of Ab 1-40: implications for the search for Ab fibril formation inhibitors. J. Struct. Biol. 130, 217–231. Soto, C., Kindy, M. S., Baumann, M. & Frangione, B. (1996). Inhibition of Alzheimer’s amyloidosis by peptides that prevent b-sheet conformation. Biochem. Biophys. Res. Commun. 226, 672–680. Soto, C. (1999). Alzheimer’s and prion disease as disorders of protein conformation: implications for the design of novel therapeutic approaches. J. Mol. Med. 77, 412–418. Permanne, B., Adessi, C. l., Saborio, G. P., Fraga, S., Frossard, M.-J., Dorpe, J. V. et al. (2002). Reduction of amyloid load and cerebral damage in a transgenic mouse model of Alzheimer’s disease by treatment with a b-sheet breaker peptide. FASEB J. 16, 860–862. Adessi, C., Frossard, M.-J., Boissard, C., Fraga, S., Bieler, S., Ruckle, T. et al. (2003). Pharmacological profiles of peptide drug candidates for the treatment of Alzheimer’s disease. J. Biol. Chem. 278, 13905– 13911. Brunger, A. T., Adams, P. D., Clore, G. M., DeLano, W. L., Gros, P., Grosse-Kunstleve, R. W. et al. (1998). Crystallography & NMR system: a new software suite for macromolecular structure determination. Acta Crystallog. sect. D, 54, 905–921. Losonczi, J., Andrec, M., Fischer, M. & Prestegard, J. (1999). Order matrix analysis of residual dipolar couplings using singular value decomposition. J. Magn. Reson. 138, 334–342. Zweckstetter, M. & Bax, A. (2000). Prediction of sterically induced alignment in a dilute liquid crystalline phase: aid to protein structure determination by NMR. J. Am. Chem. Soc. 122, 3791–3792. Prestegard, J., Bougault, C. & Kishore, A. (2004). Residual dipolar couplings in structure determination of biomolecules. Chem. Rev. 104, 3519–3540. Tolman, J., Al-Hashimi, H., Kay, L. & Prestegard, J. (2001). Structural and dynamic analysis of residual dipolar coupling data for proteins. J. Am. Chem. Soc. 123, 1416–1424. Wouters, W. A. & Curmi, P. M. G. (1995). An anaylsis of side-chain interactions and pair correlations within antiparallel beta-sheets: the differences between backbone hydrogen-bonded and non-hydrogenbonded residue pairs. Proteins: Struct. Funct. Genet. 22, 119–131. Tjernberg, L. O., Tjernberg, A., Bark, N., Shi, Y., Ruzsicska, B. P., Bu, Z. et al. (2002). Assembling amyloid fibrils from designed structures containing a significant amyloid b-peptide fragment. Biochem. J. 366, 343–351.

776


44. Balbach, J. J., Ishii, Y., Antzutkin, O. N., Leapman, R. D., Rizzo, N. W., Dyda, F. et al. (2000). Amyloid fibril formation by Ab16-22, a seven-residue fragment of the Alzheimer’s b-amyloid peptide, and structural characterization by solid state NMR. Biochemistry, 39, 13748–13759. 45. Petkova, A. T., Buntkowsky, G., Dyda, F., Leapman, R. D., Yau, W. M. & Tycko, R. (2004). Solid state NMR reveals a pH-dependent antiparallel beta-sheet registry in fibrils formed by a beta-amyloid peptide. J. Mol. Biol. 335, 247–260. 46. Stewart, R. E. & Thornton, J. M. (2002). Predictions of strand pairing in antiparallel and parallel beta-sheets using information theory. Proteins: Struct. Funct. Genet. 48, 178–191. 47. Glaubitz, C., Burnett, I. J., Grobner, G., Mason, A. J. & Watts, A. (1999). Deuterium-MAS NMR spectroscopy on oriented membrane proteins: applications to photointermediates of bacteriorhodopsin. J. Am. Chem. Soc. 121, 5787–5794. 48. Moltke, S., Nevzorov, A. A., Sakai, N., Wallat, I., Job, C., Nakanishi, K. et al. (1998). Chromophore orientation in bacteriorhodopsin determined from the angular dependence of deuterium nuclear magnetic resonance spectra of oriented purple membranes. Biochemistry, 37, 11821–11835. 49. Reif, B., Hohwy, M., Jaroniec, C. P., Rienstra, C. M. & Griffin, R. G. (2000). NH-NH vector correlation in peptides by solid state NMR. J. Magn. Reson. 145, 132– 141. 50. Ban, T., Hamada, D., Hasegawa, K., Naiki, H. & Goto, Y. (2003). Direct observation of amyloid fibril growth monitored by thioflavin T fluorescence. J. Biol. Chem. 278, 16462–16465. 51. Ban, T., Hoshino, M., Takahashi, S., Hamada, D.,

52.

53. 54.

55. 56. 57.

58.

59. 60.

Hasegawa, K., Naiki, H. & Goto, Y. (2004). Direct observation of Abeta amyloid fibril growth and inhibition. J. Mol. Biol. 334, 757–767. Lockhart, A., Ye, L., Judd, D. B., Merritt, A. T., Lowe, P. N., Morgenstern, J. L. et al. (2005). Evidence for the presence of three distinct binding sites for the thioflavin T class of Alzheimer’s disease PET imaging agents on beta-amyloid peptide fibrils. J. Biol. Chem. 280, 7677–7684. Levine, H. (2005). Multiple ligand binding sites of Abeta(1-40) fibrils. Amyloid, 12, 5–14. Klunk, W. E., Pettegrew, J. W. & Abraham, D. J. (1989). Quantitative evaluation of congo red binding to amyloid-like proteins with a beta-pleated sheet conformation. J. Histochem. Cytochem. 37, 1273. Carter, D. B. & Chou, K.-C. (1998). A model for structure-dependent binding of Congo red to Alzheimer b-amyloid fibrils. Neurobiol. Ageing, 19, 37–40. Fields, G. B. & Noble, R. L. (1990). Solid phase peptide synthesis utilizing 9-fluorenylmethoxycarbonyl amino acids. Int. J. Pept. Protein Res. 35, 161–214. Zagorski, M. G., Yang, J., Shao, H., Ma, K., Zeng, H. & Hong, A. (1999). Methodological and chemical factors affecting amyloid b peptide amyloidogenicitiy. Methods Enzymol. 309, 189–204. Eisenmesser, E. Z., Zabell, A. P. R. & Post, C. B. (2000). Accuracy of bound peptide structures determined by exchange transferred nuclear Overhauser data: a simulation study. J. Biomol. NMR, 17, 17–32. Humphrey, W., Dalke, A. & Schulten, K. (1996). VMD – visual molecular dynamics. J. Mol. Graph. 14, 33–38. Laskowski, R., MacArthur, M., Moss, D. & Thornton, J. (1993). PROCHECK: a program to check the stereochemical quality of protein structures. J. Appl. Crystallog. 26, 283–291.

Edited by A. G. Palmer III (Received 2 March 2005; received in revised form 30 August 2005; accepted 16 September 2005) Available online 5 October 2005