Report

Self-Propagating, Molecular-Level Polymorphism in Alzheimer's ß-Amyloid Fibrils

See allHide authors and affiliations

Science  14 Jan 2005:
Vol. 307, Issue 5707, pp. 262-265
DOI: 10.1126/science.1105850

Abstract

Amyloid fibrils commonly exhibit multiple distinct morphologies in electron microscope and atomic force microscope images, often within a single image field. By using electron microscopy and solid-state nuclear magnetic resonance measurements on fibrils formed by the 40-residue β-amyloid peptide of Alzheimer's disease (Aβ1–40), we show that different fibril morphologies have different underlying molecular structures, that the predominant structure can be controlled by subtle variations in fibril growth conditions, and that both morphology and molecular structure are self-propagating when fibrils grow from preformed seeds. Different Aβ1–40 fibril morphologies also have significantly different toxicities in neuronal cell cultures. These results have implications for the mechanism of amyloid formation, the phenomenon of strains in prion diseases, the role of amyloid fibrils in amyloid diseases, and the development of amyloid-based nano-materials.

Amyloid fibrils are self-assembled filamentous aggregates formed by peptides and proteins with diverse amino acid sequences (1). Current interest in amyloid fibrils arises from their involvement in Alzheimer's disease (AD), type 2 diabetes, prion diseases, and other protein misfolding disorders (2) and from basic questions about the interactions that stabilize amyloid structures and the mechanisms by which they form (3). Recent experiments additionally suggest that amyloid structures may be a basis for one-dimensional nanomaterials with possible technological applications (4, 5).

In transmission electron microscope (TEM) and atomic force microscope (AFM) images, amyloid fibrils commonly exhibit multiple distinct morphologies, often described as twisted or parallel assemblies of finer protofilaments (68). Two explanations for amyloid polymorphism are possible: (i) distinct morphologies result from distinct modes of lateral association of protofilaments without significant variations in molecular structure (7) or (ii) distinct morphologies result from significant variations in molecular structure at the protofilament level. Here, we report electron microscopy and solid-state nuclear magnetic resonance (NMR) data on amyloid fibrils formed by the 40-residue β-amyloid peptide associated with AD (Aβ1–40) that support the second possibility and reveal specific molecular-level structural differences between different fibril morphologies. The predominant morphology and molecular structure are sensitive to subtle differences in fibril growth conditions in de novo preparations (at fixed pH, temperature, buffer composition, and peptide concentration), but both morphology and molecular structure are self-propagating in seeded preparations. Different Aβ1–40 fibril morphologies also exhibit significantly different toxicities in neuronal cell cultures.

Amyloid fibrils were prepared from the human Aβ1–40 peptide (sequence NH2-DAEFRHDSGY EVHHQKLVFF AEDVGSNKGA IIGLMVGGVV-COOH) (9, 10). Parent fibrils were grown by incubation of identical fresh Aβ1–40 solutions for periods of 21 to 68 days either in vertical dialysis tubes in an unstirred bath of buffer or in horizontal polypropylene tubes with gentle circular agitation (Red Rotor orbital mixer, Hoefer Scientific Instruments, San Francisco, CA), producing quiescent and agitated parent fibrils, respectively. Daughter fibrils were grown under dialysis conditions by seeding fresh solutions with sonicated fragments (9.1% of total Aβ1–40) of either quiescent or agitated parents. Daughter fibrils were used as seeds for granddaughter fibrils, also grown under dialysis conditions. Daughter and granddaughter fibrils were grown for 3 to 8 days.

TEM images of negatively stained Aβ1–40 fibrils are shown in Fig. 1 (9). The predominant morphology for quiescent parents is a fibril with a periodically modulated width (50- to 200-nm period and 12- ± 1-nm maximum width), commonly ascribed to a periodic twist (68). The predominant morphology for agitated parents is a filament with no resolvable twist (5.5- ± 0.5-nm width) but with a pronounced tendency to associate laterally into dimers or multimers. Morphological differences are preserved in TEM images of sonicated seeds and are transmitted to daughter and granddaughter fibrils, even though all daughter and granddaughter fibrils were grown under identical dialysis conditions. AFM images show similar morphological differences (fig. S1).

Fig. 1.

TEM images of amyloid fibrils formed by the Aβ1–40 peptide, negatively stained with uranyl acetate. Parent fibrils were prepared by incubation of Aβ1–40 solutions either under quiescent dialysis conditions or in a closed polypropylene tube with gentle agitation. Daughter and granddaughter fibrils were grown under dialysis from solutions that were seeded with sonicated fragments of parent and daughter fibrils, respectively.

Molecular and supramolecular structures of amyloid fibrils can be probed by various solid-state NMR techniques (3, 1116). Figure 2 shows two-dimensional (2D) solid-state 13C NMR spectra of parent, daughter, and granddaughter fibrils with uniform labeling of F20, D23, V24, K28, G29, A30, and I31 (9). The patterns of strong one-bond cross peaks in these spectra are determined by isotropic 13C chemical shifts, which are sensitive to local structural and conformational environment (13, 17, 18). Remarkably, the 2D spectra of quiescent and agitated parent fibrils show pronounced differences in cross-peak patterns that are transmitted to the daughter and granddaughter fibrils. Thus, the morphological differences in Fig. 1 correlate with underlying differences in structure at the molecular level. The 13C chemical shift differences are greater than 1.5 parts per million (ppm) at multiple backbone and side-chain sites (tables S1 and S2) and are independent of the protocol used to prepare Aβ1–40 solutions for parent fibril growth (fig. S2).

Fig. 2.

Two-dimensional solid-state 13C NMR spectra of Aβ1–40 fibrils, prepared with 13C labeling of all carbon sites in amino acid residues F20, D23, V24, K28, G29, A30, and I31. Spectra were recorded under magic angle spinning. Regions enclosed by ellipses contain Cγ1/Cδ cross peaks of I31 (red), Cβ/Cγ cross peaks of V24 (blue), and Cα/Cβ cross peaks of F20, D23, V24, K28, and I31 (green).

The propensities for lateral association exhibited by quiescent and agitated fibrils suggest that different amino acid side chains are exposed on the fibril surfaces, possibly leading to different biological activities. Figure 3 shows the results of measurements of Aβ1–40 fibril toxicity in cultures of primary rat embryonic hippocampal neurons (9). Both quiescent and agitated fibrils are neurotoxic at Aβ1–40 concentrations of 10 μM and above, but the toxicity of quiescent fibrils is significantly higher than that of agitated fibrils. Toxicity measurements on nonfibrillar Aβ1–40 aggregates, which form in unseeded solutions at early stages of incubation, show no significant difference between quiescent and agitated conditions (fig. S3).

Fig. 3.

Toxicity of Aβ1–40 fibrils in cultures of primary embryonic rat hippocampal neurons, assayed by counting viable neurons after 24-hour and 48-hour exposures to Aβ1–40 at indicated concentrations in neurobasal medium. Error bars indicate standard errors on mean survival values (four culture dishes per condition and 35 to 145 neurons monitored per dish). Control is neurobasal medium alone. Vehicle is supernatant after centrifugation of Aβ1–40 fibril solution, added to neurobasal medium in volume equal to that for 30 μM conditions. The asterisks and plus symbols indicate conditions where toxicity differences between quiescent and agitated fibrils are statistically significant (analysis of variance test, P < 0.01 and P < 0.05, respectively).

Some specific structural features of quiescent and agitated Aβ1–40 fibrils are indicated in Fig. 4. All amyloid structures contain the cross-β motif, i.e., β sheets extending over the length of the fibril, with β strands roughly perpendicular to and interstrand hydrogen bonds roughly parallel to the long fibril axis (1, 3). Intermolecular 13C-13C nuclear magnetic dipole-dipole couplings (Fig. 4A) indicate intermolecular distances of 0.55 ± 0.05 nm at V12, V39, and A30, implying in-register, parallel β sheets for both morphologies as reported previously for Aβ10–35 and Aβ1–40 fibrils prepared with different protocols (12, 14, 15, 19). The 15N-13C dipole-dipole couplings (Fig. 4B) indicate a 0.32 ± 0.02-nm distance between side-chain Cγ carbons of D23 residues and side-chain Nζ nitrogens of K28 residues in agitated Aβ1–40 fibrils, consistent with salt bridges between oppositely charged D23 and K28 side chains. Weaker D23-K28 couplings are observed in quiescent fibrils, consistent with longer average Cγ-Nζ distances. Dipole-dipole couplings between side-chain Cδ carbons of E22 residues and side-chain Nζ nitrogens of K16 residues are observed in quiescent fibrils, possibly indicating partial occupation of an intermolecular K16-E22 salt bridge, but are absent in agitated fibrils. Histograms of fibril mass-per-length (MPL) (Fig. 4C) obtained from scanning transmission electron microscope (STEM) images (9) show mode values of 21.4 kD/nm for agitated fibrils [similar to earlier Aβ1–40 and Aβ1–42 fibril STEM data (6, 15)] and 30.3 kD/nm for quiescent fibrils. Given the 0.47- to 0.48-nm spacing between peptide chains in an ideal cross-β structure, the MPL of one layer of Aβ1–40 molecules (4.3 kD mass) would be 9.1 kD/nm. Figure 4C suggests that the protofilament (defined here as the experimentally detected structure with minimal MPL) contains two molecular layers in agitated Aβ1–40 fibrils and three molecular layers in quiescent Aβ1–40 fibrils. Deviations from precise integer multiples of 9.1 kD/nm may arise from a nonzero angle θ between the interstrand hydrogen bonding direction and the long fibril axis, which would increase the MPL values by 1/cos θ.

Fig. 4.

Molecular structural features of quiescent (blue) and agitated (red) Aβ1–40 fibrils. (A) Measurements of intermolecular 13C-13C nuclear magnetic dipole-dipole couplings for V12 carbonyl, A30 methyl, and V39 carbonyl carbons (circles, triangles, and squares, respectively) and numerical simulations for the indicated 13C-13C distances. Data were obtained for singly 13C-labeled samples with the constant-time, finite-pulse, radio frequency–driven recoupling solid-state NMR technique (9, 14). (B) Measurements of 15N-13C dipole-dipole couplings for D23 Cγ/K28 Nζ and E22 Cδ/K16 Nζ pairs (triangles and circles, respectively) and numerical simulations for the indicated 15N-13C distances. Data were obtained for samples with uniformly 15N- and 13C-labeled residues with the frequency-selective rotational echo double resonance solid-state NMR technique (9, 32). Error bars are calculated from the root mean square noise in the NMR spectra. (C) MPL histograms extracted from STEM images of Aβ1–40 fibrils. Dashed lines indicate MPL values for ideal cross-β structures with between one and five layers of Aβ1–40 molecules. a.u., arbitrary units.

Established correlations between 13C NMR chemical shifts and secondary structure (13, 17) and predictions of backbone [phis] and ψ torsion angles from the TALOS program (18) suggest that the β strand segments in quiescent Aβ1–40 fibrils include residues 10 to 14, 16 to 22, 30 to 32, and 34 to 36, whereas those in agitated Aβ1–40 fibrils include residues 10 to 22, 30 to 32, and 34 to 36 (tables S1 and S2). Figure 4A shows that V12 and V39 participate in the parallel, hydrogen-bonded structure in both polymorphs. The 13C chemical shifts for Q15, D23, G25, and G33 in quiescent fibrils and for D23, G25, and G33 in agitated fibrils suggest non-β strand conformations at these residues. The 13C NMR line-widths (fig. S4) indicate structurally disordered N-terminal segments in both morphologies, as previously suggested (13, 14, 1921). Line-widths for most CO, Cα, and Cβ sites in residues 12 to 39 of quiescent fibrils and residues 10 to 39 of agitated fibrils are 2.5 ppm or less, indicating overall structural order (13).

Data for agitated Aβ1–40 fibrils are largely consistent with our recent model for the molecular structure of the Aβ1–40 protofilament (13, 15). STEM, 15N-13C coupling, and chemical shift data for quiescent Aβ1–40 fibrils indicate a qualitatively different structure in both molecular conformation and supramolecular organization. The largest chemical shift differences, suggesting the largest conformational differences, occur at Q15 and in residues 22 to 28. Although all residues in agitated Aβ1–40 fibril samples exhibit one major set of 13C chemical shifts (with the exception of D23, V24, and K28, possibly indicating the coexistence of two distinct D23-K28 salt bridge geometries), many residues in quiescent fibril samples exhibit two sets of 13C chemical shifts, with an approximate 2:1 ratio of NMR signal intensities (e.g., I31 side-chain signals in Fig. 2). This observation raises the possibility that the quiescent Aβ1–40 protofilament contains two structurally equivalent and one structurally inequivalent subunits, consistent with Fig. 4, B and C.

We point out four physical and biological implications of these data. First, the sensitivity of fibril morphology and molecular structure to growth conditions shows that at least two distinct fibril nucleation mechanisms exist for Aβ1–40. One mechanism, leading to quiescent fibrils, may be purely homogeneous. The other mechanism, leading to agitated fibrils, may depend on the interface between the peptide solution and the air or the walls of the sample tube. The molecular structure of Aβ1–40 fibrils is not determined solely by amino acid sequence and is not purely under thermodynamic control. Second, the phenomenon of strains in prion diseases, in which a single prion protein gives rise to multiple, distinct phenotypes, has been attributed to an ability of both mammalian and yeast prion proteins to adopt multiple, distinct amyloid-like structures. Observed differences in proteolysis patterns (22, 23), resistance to chemical denaturation (24), seeding efficiencies (25), and electron paramagnetic resonance signals (26) support this proposal, but clear connections between morphological variations and molecular-level structural variations, between strains and morphological variations, and between strains and specific features of molecular structure have not yet been established experimentally. The correlations of amyloid fibril morphology with specific structural features established by our data, the demonstration of their self-propagating nature, and the observation of different neurotoxicities for different morphologies further strengthen the case for a structural origin of prion strains. Third, the importance of mature amyloid fibrils as etiological agents in AD and other amyloid diseases, as opposed to nonfibrillar oligomers observed at earlier stages of peptide incubation (2730), is a subject of current controversy. One principal argument against a primary role for mature fibrils in AD has been the absence of a robust correlation between the severity of neurological impairment and the extent of amyloid deposition (2, 31). Data in Fig. 3 raise the possibility that certain amyloid morphologies may be more pathogenic than others in the affected organs of amyloid diseases, which would weaken the correlation between disease symptoms and total amyloid deposition. Fourth, amyloid fibrils may prove useful as structural and chemical templates for self-assembled, one-dimensional nanomaterials with novel electronic or optical properties (4, 5). Because structural uniformity is a likely prerequisite in such applications, the self-propagation of molecular structure demonstrated above may be important for reliable fabrication of amyloid-based nanomaterials.

Supporting Online Material

www.sciencemag.org/cgi/content/full/307/5707/262/DC1

Materials and Methods

Figs. S1 to S4

Tables S1 and S2

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article