The Alzheimer’s Amyloid-β(1–42) Peptide Forms Off-Pathway Oligomers and Fibrils that are Distinguished Structurally by Intermolecular Organization

Abstract

Increasing evidence suggests that soluble aggregates of amyloid-β (Aβ) initiate the neurotoxicity that eventually leads to dementia in Alzheimer’s disease. Knowledge of soluble aggregate structures will enhance our understanding of the relationship between structures and toxicities. Our group has reported a stable and homogeneous preparation of Aβ(1–42) oligomers that has been characterized by various biophysical techniques. Here, we have further analyzed this species by solid state nuclear magnetic resonance (NMR) spectroscopy and compared NMR results to similar observations on amyloid fibrils. NMR experiments on Aβ(1–42) oligomers reveal chemical shifts of labeled residues that are indicative of β-strand secondary structure. Results from 2D DARR experiments indicate proximities between I31 aliphatic and F19 aromatic carbons. An isotope dilution experiment further indicates that these contacts between F19 and I31 are intermolecular, contrary to models of Aβ oligomers proposed previously by others. For Aβ(1–42) fibrils we observed similar NMR lineshapes and inter-sidechain contacts, indicating similar secondary and quaternary structures. The most prominent structural differences between Aβ(1–42) oligomers and fibrils were observed through measurements of inter-molecular ¹³C-¹³C dipolar couplings observed in PITHIRDS-CT experiments. PITHIRDS-CT data indicate that, unlike fibrils, oligomers are not characterized by in-register parallel β-sheets. Structural similarities and differences between Aβ(1–42) oligomers and fibrils suggest that folded β-strand peptide conformations form early in the course of self-assembly, and that oligomers and fibrils differ primarily in schemes of inter-molecular organization. Distinct inter-molecular arrangements between Aβ(1–42) oligomers and fibrils may explain why this oligomeric state appears off-pathway for monomer self-assembly to fibrils.

INTRODUCTION

Extracellular accumulation of aggregated amyloid-β (Aβ) peptide in neocortical regions of the brain is the pathological hallmark of Alzheimer’s disease (AD).¹ Although Aβ peptides of various lengths have been detected, Aβ(1–40) and Aβ(1–42) are the two dominant species and result from a sequential cleavage of a membrane-bound amyloid precursor protein (APP) by β- and γ-secretase.^2,3 The original amyloid hypothesis proposed that deposition of insoluble Aβ fibrils in the form of amyloid plaques induces the neuropathology which leads to AD,^4,5 but this hypothesis is unable to clearly explain why some individuals with high histological plaque counts are cognitively normal.⁶ Nevertheless, a number of genetic mutations in the APP gene promote increased production and aggregation of Aβ(1–42) and lead to early-onset familial AD.^7–9 This observation prompted the revised amyloid hypothesis that soluble, neurotoxic Aβ aggregates cause synaptic dysfunction and initiate a cascade of events that result in AD. Soluble Aβ species commonly denoted oligomers have been shown to exist in the brain extracts of AD patients,^6,10 in transgenic Tg2576 mice,^11,12 and in the conditioned medium of Chinese hamster ovary cells (7PA2 cells)^13,14 that express human APP. These oligomers can inhibit long-term potentiation (LTP), enhance long-term depression (LTD), and reduce dendritic spine density in normal rodent hippocampal slices as well as impair the memory of a learned behavior in normal rats.^10,13–15 Therefore, structural information about Aβ oligomers may provide insights into neurotoxicity that identify new drug targets or diagnostic tools for early prevention and treatment of AD.

The levels of endogenous Aβ oligomers observed to date have been too low for detailed structural studies, prompting many research groups to generate soluble Aβ aggregates from synthetic peptides. This approach has worked well for the characterization of insoluble Aβ fibrils produced from synthetic peptides, and several closely related fibril structural models that have been obtained by solid state nuclear magnetic resonance (NMR) spectroscopy analyses have been reported.^16,17 Many soluble Aβ aggregates also have been prepared from synthetic peptides, including amyloid-derived diffusible ligands (ADDLs),¹⁸ globulomers,^19–21 amylospheroids,²² oligomers prepared by photo-induced crosslinking of unmodified protein (PICUP),²³ annular protofibrils,²⁴ and protofibrils.²⁵ Several of these preparations are heterogeneous, and it has been far more difficult to isolate individual oligomers than to isolate homogeneous samples of insoluble fibrils. Based on electron microscopy (EM) and atomic force microscopy (AFM), the sizes and morphologies of these synthetic aggregates vary from small globular structures^18–20 to short fibrillar rods,²⁵ and their reported molecular weights (MWs) vary from 17 kDa for ADDLs¹⁸ to 250 х 10³ kDa for elongated protofibrils.²⁵ Nevertheless, based on circular dichroism (CD) spectra, the general consensus is that virtually all aggregated Aβ species display β structure.

The inability of some of these synthetic oligomers to proceed to fibrils raises a central structural question: how can individual peptide molecules, mostly in β-strand conformations, arrange themselves to form soluble oligomeric structures that do not readily convert to fibrils? Despite the tendency of some in vitro generated soluble Aβ oligomers to undergo further aggregation at the high concentrations required for NMR, a few structures have recently been proposed.^26–28 Spherical Aβ(1–40) aggregates denoted I_β, with MWs of 650 kDa or higher, were trapped by flash-freezing and immediate lyophilization. Based on ¹³C chemical shift, dihedral angle, and ¹³C–¹³C dipolar coupling solid state NMR analyses, no structural differences between samples of I_β and Aβ(1–40) fibrils were detected.²⁶ In particular, Aβ(1–40) molecules within fibrils are believed to adopt folded β-strand conformations that arrange to form in-register parallel β-sheets (Fig. 1A). In a solution state NMR study, Olejniczak and coworkers²⁷ reported a dimeric structure for a soluble Aβ(1–42) species, denoted a preglobulomer, in 0.2% (7 mM) sodium dodecylsulfate (SDS). In contrast to a fibril-like structure, the proposed preglobulomer structure consists of a pair of in-register parallel β-strands formed by residues near the C-terminus (L34–V40) extending from an intramolecular hairpin formed by antiparallel arrangement of two additional β-strand regions (V18–D23 and K28–G33)²⁷ (Fig. 1B). In another study, Ahmed et. al.²⁸ prepared apparent disc-shaped Aβ(1–42) pentamers by incubating Aβ(1–42) monomers under low temperature and low salt conditions. Based on amide hydrogen-deuterium exchange solution state NMR analysis and more direct solid state NMR measurements, the proposed structure contains an antiparallel arrangement of three β-strands with turn regions at H13-Q15, G25-G29, and G37-G38²⁸ (Fig. 1C). The Aβ(1–42) structural models suggest that there may be an energetic barrier preventing ready conversion of Aβ(1–42) oligomers to fibrils, imposed by distinct patterns of arrangements between β-strands. This concept is consistent with recent findings of Laganowsky et. al.²⁹ on a segment of αB crystallin, which can form an oligomer composed of an anti-parallel β-sheet or a fibril composed of parallel β-sheets. In addition, previous measurements based on Fourier transform infrared spectroscopy have suggested that β-strand organization transforms significantly during early stages of amyloid formation,^30,31 and that some preparations of Aβ(1–42) oligomers may be composed of antiparallel β-sheets.^30,31

Fig. 1.

Proposed schematic NMR models of Aβ fibrils and oligomers, illustrating differences in inter-sidechain proximities. (A) The basic structural unit of Aβ(1–40) fibril models^16,17,32,33 derived from solid state NMR. Molecules can be folded such that even or odd residues can be oriented into the hydrophobic core for either N-terminal or C-terminal β-sheets, as illustrated in the right hand side of (A), where the fiber axis is oriented perpendicular to the plane of the paper. (B) A model for dimeric structure of preglobulomers, based on solution state NMR.²⁷ (C) A model for pentameric disc-shaped oligomers, derived from amide hydrogen-deuterium exchange (solution state NMR) and solid state NMR data.²⁸

Our group recently reported synthetic Aβ(1–42) oligomers that were generated on the anionic interfaces of SDS micelles^20,21 by procedures similar to those used to obtain globulomers.¹⁹ We show here that our oligomer preparations are very stable and can be isolated by size exclusion chromatography (SEC) or further concentrated by centrifugal filtration without significant change in their biophysical features. We also present preliminary findings on solid state NMR characterization of our synthetic Aβ(1–42) oligomers. The results are discussed with relevance to proposed NMR models of Aβ oligomers^27,28 and fibrils.^16,17,32,33 Our solid state NMR data indicate that Aβ(1–42) fibrils and oligomers are both characterized by β-strand molecular conformations but differ in their inter-molecular packing within β-sheets. Specifically, 2-dimensional (2D) finite pulse radio frequency driven recoupling (fpRFDR) NMR spectra on isotope-labeled fibrils and oligomers are consistent with β-strand conformations for molecules in both samples, with similar levels of conformational heterogeneity. Inter-sidechain contacts were observed via 2D dipolar-assisted rotational resonance (DARR) between F19, I31, and A30 in both samples, consistent with folded β-strand conformation.^34,35 The primary difference between fibrils and oligomers was observed in PITHIRDS-CT based measurements of inter-molecular ¹³C-¹³C dipolar coupling.³⁶ For fibrils, PITHIRDS-CT data are consistent with the presence of in-register parallel β-sheets. PITHIRDS-CT data on oligomers are not consistent with this configuration.

RESULTS

Integrity of Aβ(1–42) oligomer preparations before and after solid state NMR.

We previously reported that Aβ(1–42) oligomers could be generated from SEC-purified Aβ monomers on SDS micelles near their critical micelle concentration.²⁰ Following dialysis to remove SDS, the oligomers were isolated by SEC. The average M of these oligomers as determined by MALS was 150 ± 30 kDa.²⁰ The SEC elution profile also showed residual monomers in this preparation that were well resolved in a later peak and constituted ~30–50% of the total Aβ.²¹ To scale up isolation of the oligomers, we concentrated the initial dialyzed sample with an Amicon Ultra 4 centrifugal concentration/filtration device (molecular weight cut-off, MWCO = 50 kDa). SEC analysis of the concentrated sample showed less than 10% monomer (Fig. 2A), and the average M of the oligomer in the eluted aggregate peak remained 152 kDa based on MALS analysis. Even with the efficient centrifugal filtration protocol for producing a concentrated oligomer with minimal contamination by monomer, at least 7 preparations were necessary to generate enough oligomer for solid state NMR analysis. Routine quality control checks confirmed highly reproducible isolation of initial Aβ(1–42) monomers (Fig. S1A) and formation of β-structure in 4 mM SDS, as monitored by the magnitude of the characteristic negative minimum at 216–218 nm in CD spectra (Fig. S1B). Quality control checks with thioflavin T also confirmed the absence of significant fibril formation during the incubation in 4 mM SDS (Fig. S1C). This fluorophore shows greatly enhanced fluorescence upon binding to amyloid fibrils,³⁷ Aβ protofibrils^25,38 and certain other Aβ aggregates enriched in β-structure.³⁹ Following dialysis and centrifugal concentration of the oligomers, both CD and thioflavin T fluorescence gave very similar results, although a small increase (by ~10–20%) in fluorescence was consistently observed among preparations (data not shown). A solution of unconcentrated oligomers stored at 4 °C typically remained stable for at least 2–3 weeks.

Fig. 2.

SEC fractionation and MALS analysis of Aβ(1–42) oligomer samples that were (A) concentrated prior to lyophilization, (B) freshly lyophilized and redissolved, and (C) recovered following solid state NMR. For SEC fractionation, an aliquot (500 μL for A and B and 1 mL of C) of sample was applied to a Superdex 75 column and eluted with 20 mM NaP buffer at 0.5 mL/min. Milli-absorbance units (mAU) were recorded at 280 nm. (A) Oligomers were concentrated with an Amicon Ultra 4 centrifugal concentration/filtration device (MWCO = 50 kDa). (B) Filtered Aβ(1–42) oligomers were immediately flash-frozen in an acetone-dry ice bath and lyophilized. The dry solid was resuspended in 20 mM NaP buffer and fractionated by SEC. (C) A small amount of the recovered Aβ(1–42) oligomer solid state NMR sample was resuspended in 20 mM NaP buffer and analyzed. Concentrations of the applied samples were 72, 100, and 100 μM for A, B, and C, respectively. The average M of the sample described in A, as determined by a simultaneous online MALS analysis, is 152 kDa. For samples B and C, the average Ms of the species inside the box are ~310 and 320 kDa, respectively, whereas the species to the left of the box (i.e. ~10% of total aggregates) are considerably larger (i.e. ~4.5 and 3.7 MDa, respectively). However, the Ms of the shoulder peak to the right of the box are ~145 and 150 kDa, respectively, consistent with the average M of our original Aβ(1–42) oligomers.²⁰ The amount of residual monomers varies from < 5% to 15% of total Aβ.

Solid state NMR measurements were routinely performed using MAS at rotation rates between 12.5 and 25 kHz in order to sharpen the observed linewidths and average away spin-spin couplings.⁴⁰ As a result, the sample can be heated between 10 °C and 30 °C higher than room temperature. Additional SEC, MALS, CD, thioflavin T fluorescence, AFM, and SDS-PAGE measurements were conducted to ensure that heat generated during the solid state NMR experiment did not significantly affect the oligomer structure. Following multidimensional solid state NMR measurements, the lyophilized oligomer sample was resuspended in 20 mM NaP buffer and compared to the initial oligomer sample prior to lyophilization and to a freshly lyophilized oligomer sample. The post-solid state NMR oligomer sample dissolved readily in buffer. Centrifugation at 18,000 × g revealed a very small pellet. Since no pellet was observed for the dissolved freshly lyophilized oligomer sample, we were concerned about potential conversion of some oligomers to fibrils. This pellet was analyzed further by thioflavin T fluorescence and AFM as described below. SEC fractionation and MALS analysis of the supernatants of the two lyophilized samples showed two distinguishable overlapping peaks for the aggregates near the void volume in addition to a small peak of residual monomers (i.e. ~ 10–15% of total aggregates) (Fig. 2B and 2C). The average M of the peak closer to the void volume (i.e. area of the peak enclosed in the box in Fig. 2B and 2C) was ~320 kDa, whereas that of the later peak was ~150 kDa, similar to that obtained previously for oligomers that had not been lyophilized. The fact that the M of the earlier peak was approximately twice that of our typical oligomers suggested a potential for further self-association of these aggregates during lyophilization. The average M of a narrow region closest to the void volume (i.e. ~10% of total aggregate peaks) was ~3.7 MDa.

Overall, results confirm that our synthetic Aβ(1–42) oligomers have good resistance to both lyophilization and heating from MAS solid state NMR measurements. CD (Fig. S2A) and thioflavin T fluorescence (Fig. S2B) analyses of both the column onput from the recovered solid state NMR sample and the fraction eluted from SEC with the highest concentration showed a minimum at 216–218 nm and low fluorescence intensity, respectively. In contrast, Aβ(1–42) monomer which had been incubated for three days under fibril-forming conditions showed a 10-fold increase in thioflavin T fluorescence intensity (Fig. S2B). Additional analysis of this column onput and selected SEC-purified fractions from the aggregate peaks by SDS-PAGE showed silver-stained bands with a distribution identical to that of the apparent decameric, trimeric, dimeric, and monomeric Aβ(1–42) species reported previously for our Aβ(1–42) oligomers²¹ (Fig. S2C). We previously concluded that the constant ratio of these bands across the observed oligomer peak suggested that our oligomers elute as a single species which partially dissociates into lower MW bands in 2% LDS loading buffer.²¹ To determine whether any oligomers converted into fibrils, AFM images for the small pellet from initial dissolution of the recovered solid state NMR sample (Fig. 3A) were compared to those for SEC-purified fraction 21 (Fig. 3B) and Aβ(1–42) fibrils (Fig. 3C). Even with rigorous vortexing, the pellet could not be dissolved in NaP buffer, but the thioflavin T fluorescence was low. Furthermore, AFM images of this sample (Fig. 3A) and the SEC-purified oligomers (Fig. 3B) displayed punctate globular structures that were very distinct from the filaments of Aβ(1–42) fibrils (Figs. 3C and S3). Thus, this pellet does not result from further aggregation of oligomers into fibrils. SDS-PAGE gel analysis of the pellet suspension revealed oligomeric bands identical to those of the SEC-purified oligomer fractions in Fig. S2C (data not shown). However, it should be noted that hydration of the lyophilized solid state NMR sample, which is commonly used to sharpen solid state NMR signals from fibrils,¹⁷ was avoided in this study. Although preliminary observations indicated little change in NMR signals, a recovered hydrated sample analyzed by SDS-PAGE showed higher MW bands and aggregates retained in the gel loading wells that indicated hydration-induced structural changes. The structural homogeneity of our oligomers is further evident in the solid state NMR results presented in sections below.

Fig. 3.

AFM images of Aβ(1–42) samples: (A) The resuspended small pellet that remained after dissolution of the solid state NMR oligomer sample; (B) SEC-purified oligomer fraction 21 (Fig. 2C, 2 μM) from the solid state NMR oligomer sample; and (C) fibrils (1 μM). The images are shown in 10 μm × 10 μm fields, with 0.4 μm × 0.8 μm insets. The AFM images of the small pellet (A) and fraction 21 (B) show globular structures, most with heights of 1–2 nm as expected for the Aβ(1–42) oligomers (20). A tiny fraction of the globular structures have heights up to 10 nm, but no protofibrils or mature fibrils are apparent. The heights of the fibrils in image C are within the range of 5–10 nm, and the lengths of individual fibrils vary between 200 nm and >1 μm. Scale bar = 1 μm.

Solid state NMR analysis of Aβ(1–42) fibrils and oligomers.

The local molecular configurations of Aβ(1–42) fibrils and oligomers were probed using the 2D fpRFDR solid state NMR technique.⁴¹ This technique creates off-diagonal peaks (crosspeaks) that connect directly ¹³C bonded atoms (i.e. distance of ~0.12nm). Figs. 4A and 4B show 2D fpRFDR solid state NMR spectra of Aβ(1–42) fibril and oligomer samples prepared with IL7 peptide, designated Samples A and B, respectively (see Table 1). The ¹³C-labeled residues in IL7 were chosen based on previously observed sensitivity of NMR observables (chemical shifts and inter-sidechain ¹³C-¹³C couplings) to different amyloid fibril molecular structures.^16,17,32,33 Analyses of peak positions (i.e. chemical shifts) and linewidths (of ~2–3 ppm), indicate that both samples were characterized by well-defined molecular structures, with most signals consistent with β-strand molecular configurations (see Table S1 and Table S2). In both samples, structural heterogeneity was visible, particularly in V24 and I31 signals: each residue exhibited more than one NMR signal per labeled ¹³C site. The less intense or minor signals are marked with colored dashed lines in Fig. 4. In contrast to the major V24 and I31 signals, the minor signals exhibited carbonyl, α-, and β-carbon ¹³C chemical shifts that are not consistent with β-strand secondary structure. It should be noted that in previous fibril structural models, V24 is believed to be within the turn region between N-terminal and C-terminal β-strands.^16,17 We also observed an additional minor crosspeak that corresponds to a crosspeak between a carbonyl- and an α-carbon in both fibril and oligomer samples (black dashed boxes in Figs. 4A and 4B). Based on longer-range transfers observed in the 2D DARR spectrum described below (Fig. 5A and 4B), we narrow the assignment of this peak to either L34 or M35 in a non-β-strand conformation (Fig. S4). Comparing all the 2D fpRFDR crosspeak lineshapes and minor signals between the fibril (Sample A) and oligomer (Sample B) samples, it would appear that both samples are characterized by similar distributions of molecular conformation. However, while lineshapes appear similar, a quantitative analysis of peak positions did not indicate equivalent structures: we calculated a root mean squared deviation of 0.65 ppm between oligomer and fibril peak positions for equivalent ¹³C-labeled sites. This degree of deviation in peak position is on the low end of the range of deviations previously reported for different Aβ(10–40) amyloid fibril polymorphs and could be consistent with structural differences.³³ Differences in oligomer and fibril 2D fpRFDR can also be seen clearly in the overlays in Fig. S5.

Fig. 4.

Two dimensional ¹³C-¹³C fpRFDR solid state NMR spectra of A) IL7 labeled fibril (Sample A) and B) IL7 labeled oligomer (Sample B). The off-diagonal crosspeaks represent single-bond correlations which allow residue-specific assignments within the ¹³C-labeled amino acids, as illustrated by colored lines. Dashed lines correspond to minor signals. The signal within the small black dashed box was observed in both fibril (carbonyl chemical shift: 176.2 ppm, C_α chemical shift: 55.3 ppm) and oligomer (carbonyl chemical shift: 176.0 ppm, C_α chemical shift: 55.0 ppm) and is assigned to either L34 or M35 (see Fig. S4).

Table 1:

Aβ(1–42) sample designations, indicating different isotopic labeling for different aggregation states.

Aβ(1–42) Sample	Aggregation State	Isotopic Labels
A	fibril	IL7: uniform 13C/15N at F19, V24, G25,A30, I31, L34, and M35
B	oligomer	IL7
C	oligomer	30% IL7; 70% unlabeled
D	fibril	13C at Cβ of A21 and carbonyl carbon (CO) of V36
E	fibril	13C at Cβ of A30 and CO of V39
F	oligomer	13C at Cβ of A21 and CO of V36
G	oligomer	13C at Cβ of A30 and CO of V39
H	monomer	IL7

Fig. 5.

Two dimensional DARR ¹³C-¹³C solid state NMR spectra of (A) fibrils (Sample A) and (B) oligomers (Sample B), showing long-range (up to ~0.6 nm) inter-sidechain contacts between F19, A30, and I31 residues at 500 ms mixing times. In the 2D DARR spectrum (A), the solid colored lines indicate single-bond exchange pathways or residue-specific assignments for F19 (black), I31 (orange), and A30 (green). In the overlaid display of horizontal slices (C), at positions indicated by horizontal dashed lines from panel A (black, at 129.4 ppm) and the 2D fpRFDR spectrum in Fig. 4A (green slice indicates the Cα peak position of A30 and orange slices indicate the Cβ and Cγ peak positions of I31), the vertical dashed lines with arrowheads indicate polarization transfers between I31 and A30 aliphatic and F19 aromatic carbons. (D) Transfers similar to those in panel C from panel B.

The molecular fold within Aβ(1–42) oligomers and fibrils was further probed using the 2D DARR technique with 500 ms mixing times. These spectra show correlations between atoms separated by distances up to ~0.6 nm (Fig. 5A and 5B). Analysis of the 2D DARR spectra from oligomer and fibril samples indicated that the F19 sidechain is near I31 sidechains in both samples. Transfers between F19 and A30 were also detected because the sequential amino acids A30 and I31 were both uniformly ¹³C-labeled. Further constraints on inter-molecular configurations were obtained by determining whether the F19-to-A30/I31 proximity corresponded to amino acids within one molecule (intramolecular) or between neighboring molecules (intermolecular). Fig. 6 shows that the F19-to- I31 contact is between neighboring molecules: analysis of a separate sample prepared with 30% IL7 and 70% unlabeled peptide (Sample C; see Table 1) showed a significant attenuation of F19-to- I31 crosspeaks. Similar results have been previously reported for Aβ fibrils^*.^16,17 SEE FOOTNOTE ^*

Fig. 6.

Solid state 2D DARR NMR spectra of the 100% IL7-labeled (Sample B, black) and 30:70% isotope-diluted Aβ(1–42) oligomer (Sample C, red) samples (500 ms mixing). Overlaid 2D DARR spectra (A) and corresponding horizontal slices (at 129.0 ppm, indicating the peak aromatic signal from F19) (B) indicate loss of F19-to-I31/A30 crosspeak intensities with isotopic dilution. The horizontal black dashed line in (A) indicates the region of the horizontal slices depicted in (B), and the vertical dashed lines in (B) indicate the signals of inter-sidechain crosspeaks which are suppressed relative to the F19 intramolecular crosspeaks due to the isotopic dilution.

Tertiary structure, defined as the alignment of β-strands within β-sheets,^40,42 was probed by PITHIRDS-CT measurement of intermolecular ¹³C–¹³C dipole-dipole coupling in oligomer and fibril samples that were selectively labeled at the CO of V39 or V36 and the C_β of A21 or A30 (Samples D-G; see Table 1). The PITHIRDS-CT decays for fibril samples (Fig. 7A) were similar to previous measurements on Aβ fibrils, with consistent decays for any backbone site within a β-sheet. Those results were the basis for assigning fibril structure to in-register parallel β-sheets.^17,43,44 In contrast, PITHIRDS-CT decays for oligomer samples (Fig. 7B) at 3 of the 4 selectively labeled sites (C_β of A21 and A30 and CO of V39) indicated that inter-molecular distances for oligomers are not consistent with in-register parallel β-sheets (> 0.5 nm). We observed a stronger oligomer PITHIRDS-CT decay for the CO of V36.

Fig. 7.

Measurements of intermolecular ¹³C-¹³C magnetic dipole–dipole couplings in C_β and CO for Aβ(1–42) fibril (A) and oligomer (B) samples by the PITHIRDS-CT solid state NMR technique. The lines represent predicted decays based on numerical simulations for linear strands of 7 ¹³C nuclei equally spaced by the specified distances. Consistent decays for fibril samples labeled at multiple sites indicate an in-register parallel β-sheet structure. Decreased decays for oligomer samples labeled at the A21 C_β, A30 C_β, and V39 CO indicate that Aβ(1–42) oligomers are not characterized by in-register parallel β-sheets.

Finally, we attempted to isolate IL7-labeled Aβ(1–42) monomers for solid state NMR analysis by conducting SEC monomer purification in 10 mM ammonium acetate, pH 8.0, followed by lyophilization (Sample H). Sample H was resuspended in NaP buffer after solid state NMR measurements, and SEC analysis (Fig. S6A) revealed that the monomer sample was unstable: about 50% of the sample had converted to aggregates. CD spectra from the recovered sample indicated some β-structure in aggregates but less than in our isolated oligomers (Fig. S6B). This structural conversion highlights the greater stability of the oligomeric structures in Samples B, C, F, and G relative to the monomers in Sample H. Solid state NMR data also were consistent with the presence of both monomers and structured aggregates, and these aggregates exhibited some features of our isolated oligomers: 2D fpRFDR spectra exhibited broad spectral features, with signal components consistent with β-strand conformations as well as significant contributions consistent with non-β-strand conformations (Fig. S7A). We also observed 2D DARR contacts between F19 and I31 (Fig. S7B). However, SDS-PAGE revealed clear differences in oligomer band distributions between recovered Sample H (Fig. S6C) and recovered oligomers (Fig. S2C). It appears that lyophilization of monomers induces some aggregation and that some intermolecular contacts in these aggregates are similar to those formed in the initial intermediate 2–4mers generated during our oligomer preparation.²⁰

DISCUSSION

The Aβ(1–42) fibril and oligomer samples analyzed here by 2D fpRFDR solid state NMR spectroscopy yield signals that are consistent with defined molecular structures and similarity in conformational distributions. For the spectra in Figs. 4 and 5, most labeled sites correspond to single NMR crosspeaks with similar linewidths for oligomer and fibril samples. Analysis of CO, C_α, and C_β secondary chemical shifts indicates that most amino acids are in β-strand conformations. Oligomers and fibrils exhibited similar crosspeak lineshapes and comparable minor structures at the same labeled sites. This degree of similarity between oligomers and fibrils is consistent with the findings of Chimon et. al., who compared structures of the I_β intermediate to Aβ(1–40) fibrils.²⁶ Observed linewidths and the existence of minor signals for V24, I31, and L34/M35 (ambiguous assignment) in both oligomer and fibril samples are consistent with previous reports of fibril polymorphism.^17,33,43 When special attention is paid to producing structurally homogenous fibrils, typical ¹³C NMR linewidths decrease to below 1 ppm and minor signals are eliminated.^17,45 Thus, all existing Aβ(1–40) fibril structural models are symmetric (all molecules are structurally equivalent), and the existence of minor NMR signals is attributed to the co-existence of fibrils with different molecular structures. One alternative interpretation of our present results is that minor signals observed in fibril samples may be due to non-fibrillar oligomeric structures which coexist with fibrils. Another possibility for our oligomers is that asymmetry creates non-equivalent Aβ(1–42) molecules with distinct molecular conformations within the same oligomeric structure. This possibility is strengthened by the consistent distribution of multiple Aβ bands on SDS-PAGE produced by partial dissociation of the oligomers in SDS.

The 2D DARR data on Aβ(1–42) oligomers and fibrils reveal that certain non-adjacent residues in the primary structure are in close proximity in the molecular fold. Specifically, the data in Fig. 5 indicate that F19 is in close proximity with I31 and A30 (within 0.6 nm) in both oligomers and fibrils. For fibril structural models, this sidechain proximity is interpreted in terms of a folded β-strand structure which includes a turn region somewhere between F19 and A30.^16,17,46 For Aβ(1–40) fibrils, contacts between F19 or F20 and odd numbered residues (I31/G33/M35) or even numbered residues (I32/L34/V36) on the C-terminal β-strand have all been observed experimentally.^16,17,32,33 The specific 2D DARR correlation between F19 and I31, as observed here, has been reported for the Aβ(10–40) fragment as well as for brain-seeded Aβ(1–40) fibrils.^32,33 In their structural model for Aβ(1–42) fibrils, Luhrs et. al. proposed that F19 is not near I31 due to significant differences in the turn region of Aβ(1–42); the present Aβ(1–42) fibril data are not consistent with this proposal.⁴⁷ When the 2D DARR experiment was previously applied to amyloid fibrils prepared with mixtures of isotopically labeled and unlabeled peptide, attenuation of inter-sidechain crosspeaks to F19 relative to F19 intra-sidechain crosspeaks indicated that sidechains on the C-terminal β-strand were near F19 sidechains on neighboring molecules. This result was rationalized in terms of a staggered alignment of β-sheets such that proximate β-strands on adjacent β-sheets are on different molecules (see Fig. 1A).^16,17 The effects of isotopic dilution on our 2D DARR oligomer data (Fig. 6B) suggest a similar configuration, although other possibilities exist in the absence of more structural constraints on oligomers. Inter-molecular proximity between non-adjacent amino acids could, for example, be consistent with anti-parallel β-sheets.

PITHIRDS-CT data indicate that Aβ(1–42) fibrils and oligomers in this study are characterized by different arrangements of adjacent β-strands. In these experiments, the decay in ¹³C NMR intensity for each selectively labeled site is dependent on positions relative to equivalent labeled sites on neighboring molecules.³⁶ For Aβ(1–42) fibrils, PITHIRDS-CT data (Fig. 7A) are consistent with in-register parallel β-sheet structure. Similar results for selectively ¹³C-labeled Aβ(1–40) and Aβ(1–42) fibrils (some based on the similar fpRFDR-CT technique⁴¹) were used by Tycko and coworkers to show that Aβ fibrils are in-register parallel β-sheets.^42,44 This inter-molecular arrangement also agrees with the model of Luhrs et. al. for Aβ(1–42) fibrils, although their model is not based on dipolar coupling measurements.⁴⁷ In contrast, our PITHIRDS-CT data on Aβ(1–42) oligomers (Fig. 7B) indicate longer ¹³C-¹³C distances than observed for the Aβ(1–42) fibrils at A21 C_β, A30 C_β, and V39 CO, indicating that our oligomers are not composed of in-register parallel β-sheets. The PITHIRDS-CT result for the V36 CO in oligomer samples reveals a shorter inter-molecular distance between equivalent sites, comparable to that in fibrils. These results are consistent with inter-molecular antiparallel β-strand interactions, perhaps similar to those observed in oligomers formed by segments of αB crystallin,²⁹ and they provide important constraints to future structural modeling. However, a close inter-molecular proximity for a single labeled site is insufficient to fully specify the inter-molecular arrangement.

The finding that oligomers differ from fibrils in inter-molecular organization agrees in a general sense with predictions of Yu et. al.²⁷ and Ahmed et. al.,²⁸ but there are discrepancies between both models and the present experimental constraints. Both models predict that oligomer structures are not consistent with fibril-like inter-molecular arrangements, in which each peptide molecule contributes one β-strand to two separate in-register parallel β-sheets (Fig. 1). The structural model of Yu et. al. predicts a parallel intermolecular arrangement between β-strands formed between residues L34 and V40, which is not consistent with our PITHIRDS-CT results for oligomers ¹³C-labeled at the V39 CO. This model is also inconsistent with our 2D DARR data, which indicate inter-molecular proximity between the F19 and I31 sidechains. Furthermore, our observed attenuation of the F19-to-I31 2D DARR contact following isotopic dilution is inconsistent with the overall topology of Yu et. al.’s model, which predicts that residues which are non-adjacent in the primary structure are brought into close proximity by an intra-molecular hairpin structure (Fig. 1B). The model of Ahmed et. al. does not predict in-register parallel β-sheets within Aβ(1–42) oligomers (Fig. 1C), but it also does not predict spatial proximity between F19 and I31 or the short inter-molecular distance we measured between V36 CO atoms. Fig. 8 depicts a hypothesized alternative configuration within Aβ(1–42) oliogmers analyzed in the present work. This configuration is consistent with the PITHIRDS-CT data in Fig. 7B, as well as the F19-I31 inter-sidechain proximity observed in Figs. 5 and 6. In this model, each molecule would contribute a single β-strand to two separate β-sheets. The β-sheets are arranged with a stagger, as defined by Petkova et. al.,¹⁶ such that closest proximities between F19 and I31 are between sidechains on different molecules. The proposed configuration is similar to the fibril structural model recently proposed by Qiang et. al. for the Iowa-mutant of Aβ(1–40), though this model does not predict a staggered β-strand arrangement between β-sheets.⁴⁸ The staggered β-strands within Aβ(1–42) oligomers may induce a β-sheet curvature which prevents the oligomer structure from extending into a fibrillar structure. Observed structural differences between present and previous oligomer preparations may be due to the effect of self-assembly conditions on Aβ structure, which is well documented for fibrils prepared under different conditions.^17,33,43 Self-assembly conditions include the purification of Aβ(1–42) monomers to remove pre-existing aggregates prior to initiating aggregation protocols. In our experience, with Aβ(1–42) stocks (~100 mg) obtained directly from peptide synthesis suppliers without HPLC purification,²⁰ SEC purification is the only way to remove residual aggregates that may constitute nearly half of the sample (Fig. S1A). Widely used protocols involving dissolution in hexafluoroisopropanol (HFIP), drying, and subsequent dissolution in DMSO or NaOH, including those in the Yu et. al.²⁷ and Ahmed et. al.²⁸ reports, do not completely remove these aggregates^49–51 (and see Fig. S1A and S8). Residual aggregates may seed or otherwise direct the pathways involved in the aggregation process.

Fig. 8.

Proposed schematic molecular configuration for Aβ(1–42) oligomers. Expected positions of sidechains are depicted by colored bars or dots, using the color scheme defined in Fig. 1. The left and right hand sides are two different views of the proposed structure. The β-sheets are staggered so that closest pairs of F19 and I31 sidechains are on different molecules. The dotted curves represent turn regions connected with β-strands that are not drawn but are expected to be within the structure.

The observed structural differences between Aβ(1–42) oligomers and fibrils may explain why Aβ(1–42) oligomers prepared by our protocol appear to be off-pathway for the conversion of monomers to fibrils.²⁰ In contrast to fibrils, the β-strands within oligomers are not organized into in-register parallel β-sheets. This distinct β-strand organization may represent an energetic barrier which prevents oligomers from readily converting into fibrils or seeding the conversion of monomers to fibrils. In other words, without the fibril-like property of each molecule contributing to two in-register parallel β-sheets, it may not be possible for oligomeric structures to extend indefinitely in one dimension. Distinct intermolecular organization of β-strands may also explain why our oligomers do not show fluorescence with thioflavin T. In contrast, the I_β Aβ(1–40) oligomers prepared by Chimon et. al. are composed of in-register parallel β-sheets and are capable of seeding fibril growth.²⁶ Despite the difference in intermolecular organization, we suggest that individual molecular conformations in our Aβ(1–42) oligomers resemble more closely the conformations of molecules within amyloid fibrils than those proposed by the models of Yu et. al. and Ahmed et. al (Figs. 1 and 8). This suggestion is based on the similar chemical shifts and lineshapes observed in the 2D fpRFDR spectra (Fig. 4), the similar 2D DARR contacts between F19 and I31 (Fig. 5), and the common attenuation of 2D DARR contacts with isotopic dilution observed for Aβ(1–42) oligomers and fibrils (Fig. 6). The distinguishing characteristic between oligomers and fibrils would then be the organization of neighboring folded β-strand molecules. If this suggestion is true, it would imply that monomeric conversion to folded β-strands precedes inter-molecular assembly in the pathway to fibril formation.

MATERIALS AND METHODS

Chemicals and Aβ Preparations.

All chemicals and solvents were purchased from Sigma-Aldrich (St. Louis, MO) and used without further purification. Aβ(1–42) with uniform ¹³C- and ¹⁵N-labeling at the amino acid residues F19, V24, G25, A30, I31, L34, and M35 (denoted IL7) was synthesized according to standard FMOC solid phase synthesis by the Peptide Synthesis Facility at the Mayo Clinic (Rochester, MN). In addition, Aβ_1–42 with more selective ¹³C isotopic labeling (i.e., ¹³C-labeling at the β-carbon (C_β) of A21 and the carbonyl carbon (CO) of V36 or at C_β of A30 and the CO of V39) was purchased from the Keck Biotechnology Resource Laboratory (New Haven, CT). Unlabeled Aβ(1–42) was obtained from both suppliers. With all of these Aβ(1–42) preparations an HPLC purification step was omitted to conserve peptide,²⁰ and their purity was determined by MALDI-mass spectrometry to be > 90%. Lyophilized Aβ(1–42) peptides, stored desiccated at −80 °C, were purified by SEC as previously described.²⁰ Briefly, monomers were isolated by dissolving 5–6 mg of crude Aβ in 35 mM NaOH. Peptide stocks that were partially insoluble in 35 mM NaOH were first dissolved in 400 μl HFIP, dried into a thin film under a gentle stream of nitrogen, and further dried by centrifugal vacuum concentration before complete dissolution in 35 mM NaOH. Dissolved samples were centrifuged at 18,000 × g for 10 min, and the supernatants were applied on a Superdex 75HR 10/30 column (Amersham Pharmacia, Piscataway, NJ) equilibrated with 20 mM sodium phosphate, pH 7.5 (NaP buffer) at a flow rate of 0.5 mL/min. Collected fractions were quantified by UV absorbance at 276 nm (with ε = 1450 M^–1cm⁻¹).²⁵ Monomer analyzed by solid state NMR was purified by SEC in 10 mM ammonium acetate, pH 8.0, and lyophilized as described in the following section.

Aβ(1–42) Oligomer and Fibril Preparations for solid state NMR.

Aβ oligomers were prepared as previously described with slight modifications.²⁰ Briefly, aliquots of 100 μM SEC-purified Aβ monomer were incubated overnight at room temperature in 50 mM NaCl and 4 mM SDS to give initial oligomers that we denote 2–4mers.²⁰ Next, the sample was dialyzed against 10 mM NaP buffer over 48 h with at least five buffer exchanges. The quality of the sample was monitored and confirmed at each step of the preparation by CD and thioflavin T fluorescence.²⁰ Residual or unconverted monomer²¹ was removed by filtering the dialyzed oligomer with an Amicon Ultra 4 centrifugal concentration/filtration device with a MW cutoff of 50 kDa. In order to prepare sufficient amounts for solid state NMR analysis (3–5mg), at least 7 consecutive samples were prepared within 1–2 days. Finally, the preparations were pooled, flash-frozen in a dry ice-acetone bath, immediately lyophilized, and stored at −40 °C until use. For isotope dilution experiments, an oligomer sample was prepared from a mixture containing 30% (v/v) isotope-labeled and 70% (v/v) unlabeled Aβ(1–42) monomers. To obtain a fibril sample, Aβ(1–42) monomer (100 μM) was incubated with 150 mM NaCl in 10 mM NaP buffer at 37 °C for 3–5 days under quiescent conditions. To speed up fibril formation and fibril homogeneity, the sample was seeded with 10% (v/v) preformed Aβ(1–42) fibrils. Formation of fibrils was monitored by an increase in fluorescence intensity resulting from thioflavin T binding.³⁷ For solid state NMR measurements, fibrils were initially pelleted by sedimentation and the supernatant discarded. The pellet was then resuspended in nano-purified water, pelleted again, flash-frozen in a dry ice-acetone bath, and lyophilized.

Multi-angle Light Scattering (MALS).

Samples were applied to a Superdex 75 SEC column attached to an AKTA FPLC system and analyzed in-line with a DAWN EOS MALS instrument using ASTRA for Windows 4.90.04 (Wyatt Technology, Santa Barbara).^25,39 The analysis was based on the Zimm formalism of the Rayleigh-Debye-Gans model,^52,53 as presented previously.²⁵ In brief, the excess Rayleigh ratio R_θ is related to the molecular structure according to eq 1,

$$\frac{K_C}{R_{\theta }}=\frac{1}{MP\left(\theta \right)}+2A_{2}C$$

where R_θ is proportional to the fraction of incident light that is scattered by the solute without interference; K is a physical constant equal to 4π²(dn/dc)²n_o²N_A⁻¹λ_o⁻⁴, where n is the refractive index of the solution, c is the solute concentration (g/mL), n_o is the refractive index of the solvent, N_A is Avagadro’s number and λ_o is the wavelength of the incident light in vacuum, A₂ is the second virial coefficient; and M is the molecular mass of the solute. At the low concentrations c employed in this study, the 2A₂c term in eq 1 may be ignored. The function P(θ) is the ratio of the scattered light intensity to the scattered light intensity without interference. For bovine serum albumin and the relatively small peptide aggregates here, R_θ showed no angular dependence and P(θ) was set to one. In calibration runs with bovine serum albumin, the monomer peak was preceded by a small shoulder of dimer, and the M for monomer was determined to be 76 ± 1 kDa (average of 4 SEC experiments).

Fluorescence spectroscopy with thioflavin T.

Fluorescence measurements were performed on a Varian Cary Eclipse fluorescence spectrophotometer (Agilent Technologies, Santa Clara, CA). Briefly, a 5-μL aliquot of aggregated 100 μM Aβ(1–42) was diluted 15 fold with a 5-μM solution of thioflavin T (in 20 mM NaP buffer) in a micro-fluorescence cuvette. The contents were gently mixed by pipeting up and down three times, and the fluorescence was measured continuously for 1 min at fixed excitation and emission wavelength (486 nm and 443 nm, respectively) with the 10nm slit widths. Background was corrected by subtracting the fluorescence of a blank sample that had been acquired under identical conditions.

Atomic Force Microscopy (AFM).

Samples were incubated for 15–30 min on freshly cleaved mica that had been modified with 3’-(aminopropyl)triethoxysilane (APTES) as previously described.⁵⁴ The mica disc was then washed with deionized water multiple times and dried overnight over dessicant before imaging. A NanoScope III controller with a Mulitmode AFM (Veeco Instruments Inc, Chadds Ford PA) was used for imaging by ambient tapping mode. Images were obtained in height mode, where increasing brightness indicates increasing feature height. Height images were “flattened,” and particle heights were collected using the “section” function of the NanoScope (version 613b21) software.

Polyacrylamide Gel Electrophoreses (PAGE).

Aβ(1–42) oligomer samples (50 pmol) in loading buffer (NuPAGE, Invitrogen Inc, Carlsbad, CA) containing 2% lithium dodecyl sulfate (LDS) were applied to NuPAGE pre-cast 4–12% acrylamide gels containing bis-Tris and resolved in NuPAGE MES SDS running buffer with 0.1% SDS. Dye-linked MW markers (SeeBlue Plus2 Prestained Standards, Invitrogen) were run in parallel for calibration. The gels were stained with silver (Pierce SilverSNAP Stain Kit II) and imaged by FluorChem (ProteinSimple, Santa Clara, CA).

Solid State NMR Spectroscopy.

Solid state NMR experiments were performed on a Bruker 11.75 Tesla (500 MHz ¹H NMR frequency) solid state NMR system, equipped with an Avance III console and a 2.5 mm magic angle spinning (MAS) NMR probe. Two-dimensional exchange NMR was performed on samples prepared with IL7 peptide. During the 2D exchange mixing periods, ¹³C-¹³C dipolar couplings eliminated by MAS were reintroduced using either fpRFDR or DARR pulse sequence, at sample-rotation rates of 25 kHz or 16.5 kHz, respectively.^34,35,41 Sample temperature was calibrated with ²⁰⁷Pb NMR of lead nitrate. Decoupling of ¹H using two pulse phase modulation was employed during free evolution and fpRFDR recoupling periods, with a ¹H radio frequency field of 110 kHz.⁵⁵ Exchange periods during 2D DARR experiments required continuous irradiation for 500 ms with ¹H fields with powers corresponding to 16.5 kHz nutation frequencies (equal to the MAS spinning rate). The π pulses on ¹³C required for fpRFDR recoupling were applied using 37.5 kHz radio frequency fields, so that the pulse durations (13.3 μs) were 1/3 of the rotor period. Each 2D fpRFDR spectrum in Fig. 4 is the result of 48 hours of signal averaging, and the 2D DARR spectra in Fig. 5 are each the result of 96 hours of signal averaging. Due to isotopic dilution, increased signal averaging (144 hours) was required for the 2D DARR spectrum of Sample C (see Fig. 6).

Samples with selective ¹³C isotopic labeling (at isolated CO or C_β sites) were analyzed using the PITHIRDS-CT experiment defined by Tycko.³⁶ This experiment was performed with MAS speeds of 12.5 kHz. PITHIRDS recoupling was applied for a total of 61.44 ms, with blocks adjusted to yield effective recoupling times between 0 and 61.44 ms (k₁ = 2, k₂ + k₃ = 32, as defined by Tycko). Continuous wave decoupling of ¹H was applied at a field of 110 kHz during PITHIRDS recoupling and 90 kHz during acquisition. Each PITHIRDS-CT curve in Fig. 7 corresponds to 22 hours of signal averaging. Data were corrected for the effects of natural abundance ¹³C background as described previously.¹⁷ Simulated PITHIRDS-CT curves were generated using Spinevolution⁵⁶ using parameters which matched the experimental conditions. Spin simulations included linear arrays of 7 ¹³C nuclear spins, separated by equal distances of 0.5 nm, 0.6 nm, 0.7 nm, or 0.8 nm. The REPULSION power averaging scheme was used for the simulations, with 168 pairs of α and β Euler angles and 12 γ angles.⁵⁷

Abbreviations:

Aβ

amyloid-β

AD

Alzheimer’s disease

ADDLS

amyloid-derived diffusible ligands

AFM

atomic force microscopy

APP

amyloid precursor protein

APTES

3’-(aminopropyl)triethoxysilane

C_β

β-carbon

CD

circular dichroism

CO

carbonyl carbon

DARR

dipolar-assisted rotational resonance

DMSO

dimethylsulfoxide

EM

electron microscopy

fpRFDR

finite pulse radio frequency driven recoupling

HFIP

hexafluoroisopropanol

HPLC

high performance liquid chromatography

LDS

lithium dodecyl sulfate

LTP

long-term potentiation

LTD

long-term depression

M

molecular mass

MALDI

matrix-assisted laser desorption/ionization

MALS

multi-angle light scattering

MAS

magic angle spinning

MW

molecular weight

MWCO

molecular weight cut-off

NMR

nuclear magnetic resonance

PICUP

photo-induced crosslinking of unmodified protein

SDS

sodium dodecylsulfate

SEC

size exclusion chromatography

2D

2-dimensional