A long-lived Aβ oligomer resistant to fibrillization

Abstract

The hydrophobic Aβ peptide is highly aggregation prone; it first forms soluble oligomers, which then convert into the amyloid fibrils found in the cerebral plaques of Alzheimer’s disease. It is generally understood that as the peptide concentration of Aβ increases, the fibrillization process is accelerated, but we examine the limits on this phenomenon. We found that once a threshold concentration of Aβ is exceeded, a stable oligomer is formed at the expense of fibril formation. The suppression of fibril formation was observed by amyloid-binding dye Thioflavin T and solution nuclear magnetic resonance (NMR). Small-angle X-ray scattering, size exclusion chromatography, and analytical ultracentrifugation demonstrated that Aβ peptides form a range of compact species, with a dimer being an early highly populated oligomer. Solution NMR allowed us to define the secondary structure of this Aβ dimer, which shows interlocking contacts between C-terminal peptide strands. Thus, we present a novel Aβ oligomer that resists conversion to fibrils and remains stable for more than one year.

Introduction

The short, hydrophobic β-amyloid (Aβ) peptide accumulates as amyloid fibrils in the plaques of Alzheimer’s disease (AD). While these insoluble plaques are a major hallmark of AD, experimental evidence suggests that soluble oligomeric complexes are more cytotoxic than fibrils ^1-2. In fact, some investigators have hypothesized that amyloids may be protective because they sequester these toxic oligomers and remove them from the environment ^3-4. On the other hand, evidence is accumulating that Aβ oligomers and fibrils have antimicrobial properties and may, in fact, serve as a primitive form of innate immunity, similar to more traditional antimicrobial peptides ^5-6. Nonfibrillar Aβ oligomers could be (1) obligate, meaning they are a necessary step in the fibril formation pathway; (2) on-pathway, meaning they are capable of converting into fibrils but not absolutely necessary for fibrillization; or (3) off-pathway, meaning they are incapable of converting into fibrils ⁷. How, and even if, these possible oligomeric complexes differ remains unknown.

In recent years, solid-state nuclear magnetic resonance (NMR) has provided structural information about Aβ amyloid fibrils formed synthetically ^8-15 or seeded by AD brain plaques ^16-17. This method has also provided insight into the secondary structure seen in different preparations of oligomers ^{1, 18-21}, including a toxic oligomer that forms a structured conformer containing parallel β-sheets ²². On-pathway oligomers are generally quickly converted into fibrils and are thus transient by nature. Previous work has suggested that addition of sodium dodecyl sulfate (SDS) or fatty acids can stabilize certain Aβ oligomers as off-pathway to fibril formation ²³, but there are no examples of these inert oligomers under standard Aβ preparation protocols.

In preparing Aβ fibril samples, some fraction of stable oligomers may be observed as a minor proportion of the total species present, but it is unclear if these oligomers are part of the fibrillization pathway ^{21, 23-24}. During the fibrillization process, as Aβ concentration increases, the formation of fibrils occurs more quickly, with lag phase inversely related to concentration ²⁵. However, mathematical modeling of amyloid formation has suggested that as the concentration exceeds a threshold, the energetic forces favor oligomerization rather than fibrillization ⁷. This phenomenon has been observed experimentally for a number of fibril-forming peptides and proteins ^26-28. Knowing the structural contacts of such an oligomer could help prevent the on-pathway oligomer structures that may be associated with toxicity.

Here, we present a high concentration Aβ oligomer that resists fibrillization detected by amyloid-binding dye Thioflavin T (ThT) and solution NMR. Using small-angle X-ray scattering (SAXS), size exclusion chromatography (SEC), and analytical ultracentrifugation, we have determined that the oligomeric state is comprised of a range of compact oligomers, with a dimer being the most prevalent species. Finally, we used solution NMR to develop a structural model for this fibril-resistant dimer.

Methods

Sample preparation

We used Aβ 1-40 (Bachem H-1194) for almost all experiments, except for 2D-NMR, in which we used ¹³C,¹⁵N uniformly labeled Aβ 1-40 (AlexoTech ABNC-240). We found peptide quality to be essential in preparing our samples, and quality control tested each peptide batch for reproducibility of fibrillization kinetics and resistance of the resulting fibrils to proteinase K digestion. We found that Aβ peptides from other commercial sources had various peptide or small molecule impurities that inhibited oligomerization.

Aβ samples were prepared by first suspending the lyophilized Aβ peptide in DMSO (10 mg/ml for low concentration samples [<185 μM, 0.8 mg/ml] and 20 mg/ml for high concentration samples [≥185 μM, 0.8 mg/ml]). The sample was then diluted to the desired final concentration in 10 mM NaPhos buffer, pH 7.2. For NMR samples, peptides were suspended in fresh DMSO-d₆ and then diluted in 10 mM NaPhos buffer, pH 7.2, containing 5% D₂O.

Thioflavin T kinetics

Aβ samples were prepared to a final concentration of 0.1–1.0 mg/ml in 10 mM NaPhos buffer, pH 7.2, and each contained the same amount of DMSO (20% final concentration). 10 μM ThT was added to each sample and then distributed in a non-binding surface 96-well plate (Corning 3991) sealed with a clear film (Nunc 235307). Fluorescence readings (λ_ex: 444 nm, λ_em: 485 nm) were taken in a SpectraMax M5 plate reader (Molecular Devices) set to 37°C with no agitation between readings.

Electron microscopy

Following 2 weeks of NMR readings, the 185 μM (0.8 mg/ml) Aβ solution was adsorbed onto freshly glow-discharged, 200-mesh formvar/carbon-coated copper grids (Ted Pella) and negative-stained with a 2% (w/v) uranyl acetate solution. Images were taken on a Philips/FEI Tecnai F20 electron microscope at 80 kV.

Small-angle X-ray scattering

Aβ samples were prepared to a final concentration of 0.2–2.0 mg/ml in 10 mM NaPhos buffer, pH 7.2, and each contained the same amount of DMSO (21.66% final concentration). The samples were left at room temperature overnight. The next day they were transferred into 1.5 mm quartz-glass capillaries (Hilgenberg GmbH, Mark-tubes, code no: 4017515), sealed, and measured with synchrotron radiation. SAXS experiments were performed at the Stanford Synchrotron Radiation Laboratory (BL 4-2) using monochromatic X-rays with energies of 10 keV. Scattered radiation was collected using a Rayonix MX225-HE CCD area detector (73.2 μm pixel size). The 2D SAXS spectra were azimuthally integrated using the Nika package plugin of Igor Pro in order to obtain the scattered intensity as a function of the magnitude of the scattering vector, I(Q) vs. Q. The scattering intensities were normalized to beam exposure as measured by an X-ray photodiode placed in the transmitted beam stop. The normalized data from the Aβ samples were then buffer subtracted and visually examined to ensure that for each sample the I(Q) converged to zero at large Q values. Samples and buffer were measured at different exposure times, and identical samples were prepared and measured to check for systematic consistency within the data set.

ANS fluorescence

The organic compound 1-anilino-8-naphthalene sulfonate (ANS) was used as a structural probe; the molecule binds to apolar regions of molten globule proteins and fluoresces when bound to an exposed hydrophobic pocket. As a control experiment, 10 μM ANS was added to 10 mg/ml bovine serum albumin (BSA) in 10 mM NaPhos buffer, pH 7.2, to confirm that the probe detects the hydrophobic pocket. Next, 10 μM ANS was incubated with 1.0 mg/ml (231 μM) Aβ in 10 mM NaPhos buffer, pH 7.2, with 20% DMSO during the oligomerization time period. ANS fluorescence (λ_ex: 370 nm, λ_em: 460 nm) was measured in a SpectraMax M5 plate reader set to 37°C with no agitation between readings.

Size-exclusion chromatography

A 1 mM Aβ sample was prepared as above; half was injected into the column immediately (“fresh” sample) and 100 μl was incubated at room temperature for two weeks and then injected (“aged” sample). The species within the Aβ sample were separated by size on a HiLoad 16/600 Superdex 75 column (GE 28989333) at a flow rate of 1.5 ml/min. [Specs: 16 × 600 mm, 120 ml bed volume, composite of cross-linked agarose and dextran, 24-44 μm particle size.] UV absorbance was collected at 220 nm.

Analytical ultracentrifugation

A fresh Aβ 1-40 peptide solution was prepared as above at a concentration of 1 mg/ml (230 µM) in 10 mM NaPhos buffer, pH 7.2, with 22% DMSO. Sedimentation velocity at 45,000 rpm (25°C) was acquired in an Optima XL-I analytical ultracentrifuge (Beckman) for 22.5 hours using interference optics every 15 minutes. Spectra were plotted in Sedfit (v. 14.3, freely available at http://www.analyticalultracentrifugation.com) to calculate the continuous c(s) distribution ^29-30. Parameters used for this calculation: buffer density = 1.0223 g cm^-3, buffer viscosity = 0.0122, partial specific volume = 0.73 (estimated from literature ^31-32).

NMR

For 1D-NMR, lyophilized Aβ peptide was suspended in fresh DMSO-d₆ (10 mg/ml for the 0.2 mg/ml sample; 20 mg/ml for higher concentration samples) and then diluted to 0.2 mg/ml, 0.8 mg/ml, or 4.3 mg/ml (1 mM) in 10 mM NaPhos buffer, pH 7.2, containing 5% D₂O. For 2D-NMR, lyophilized ¹³C,¹⁵N uniformly labeled Aβ peptide was suspended in fresh DMSO-d₆ (20 mg/ml) and then diluted to 4.3 mg/ml (1 mM) in 10 mM NaPhos buffer, pH 7.2, containing 5% D₂O.

NMR methods

NMR spectra were recorded at 298K and 285K on a Bruker Avance II 900 MHz spectrometer equipped with cryogenic probe. At 298K, 3D HNCACB/ CBCA(CO)NH and 3D HNCO/ CO(CA)NH were recorded to obtain sequence specific backbone (¹H^N, ¹⁵N, ¹³C^α, ¹³CO) and ¹³C^β resonance assignments. ¹H^α and ¹H^β assignments were extended by the 3D HAHB(CO)NH experiment, and more peripheral sidechain chemical shifts were assigned with aliphatic 3D CCH-TOCSY (mixing time: 75 ms). 3D ¹⁵N-and ¹³C^aliphatic-resolved [¹H,¹H]-NOESY (mixing time: 120 ms) were recorded along with 2D high-resolution ¹H-¹H NOESY.

Because the NMR samples were prepared from a concentrated stock solution in DMSO, and the concentration of DMSO was significant at the highest concentration (Methods), we were concerned that the solvent might influence the conformation of the protein. We therefore compared the NMR spectra of two different peptide concentrations: 185 μM and 1 mM. Both are above the threshold to form the stable oligomer, but the 185 μM sample contains only 4% volume DMSO while the 1 mM sample requires 21% volume DMSO to properly suspend the peptide prior to sample preparation. The ¹³C-HSQC NMR spectra for these two samples are nearly identical (Fig S3), demonstrating that neither concentration nor a high percentage of DMSO alters the structure of the oligomer.

Results

Very high concentrations of Aβ suppress fibril formation

While it is well accepted in the field that increasing the concentration of Aβ reduces the lag phase to fibril formation ^{25, 33}, we wanted to explore the limits of this trend. We first used the amyloid-binding dye ThT to follow the kinetics of Aβ 1-40 (Aβ40) fibrillization at different concentrations. Agitating the system accelerates fibril formation by shearing fibrils and promoting self-seeding, so we investigated fibril formation without shaking to minimize extrinsic influences on fibril formation. As convention dictates, we noticed that increasing the concentration of Aβ from 0.1 mg/ml (23 μM) to 0.2 mg/ml (46 μM) and to 0.4 mg/ml (92 μM) resulted in more rapid fibril formation, with shorter lag phases that decreased as Aβ concentration increased (Fig 1A). However, as we increased the concentration further to 0.8 mg/ml (185 μM) and 1.0 mg/ml (231 μM), we observed that the lag phase was actually extended and fibril formation was suppressed (Fig 1A). By plotting the average time to reach half the maximum signal (t_1/2) (Fig 1A, right panel), it is clear that the curve reaches a maximum, with fibril formation accelerating with concentration up to 185 μM and then decelerating again as concentration exceeds this threshold.

Figure 1.

Increasing concentration of Aβ suppresses fibril formation. A. Unagitated fibril formation at the indicated Aβ concentration (0.1–1 mg/ml) was monitored by amyloid-binding dye ThT (10 μM) for 2 weeks. In the panel on the right, the t_1/2 time was plotted to show average lag time for each sample. For samples that did not form fibrils during the observation period, a lag phase of 14 d was assumed for these calculations. B. Normalized peak intensity of the methyl groups as measured by NMR was plotted over time, allowing us to detect decay of signal as large NMR-invisible conformations accumulated. While the lower concentration (46 μM) eventually disappeared completely, indicating that all soluble species were consumed in the process of fibrillization, the signals of the higher concentrations (185 μM, 1 mM) persisted for more than two weeks. C. Following the 2-week reading by NMR in panel B, the 185 μM oligomer sample was examined by electron microscopy (EM). Negative stain EM showed that some fibrils were present, but not nearly at the density expected for this concentration, suggesting that the remaining Aβ was not seeded by these fibrils. Scale bar = 100 nm.

To confirm that delayed fibrillization was observed by other methods that did not rely on ThT as an extraneous probe, we used NMR to monitor the disappearance of Aβ40 from solution. To do this, we recorded the 1D-NMR spectrum of naturally occurring ¹H. As large, insoluble aggregates such as fibrils are formed, the proteins become invisible to NMR and the intensity of the peaks decreases; thus, a disappearance of signal indicates fibril formation ³⁴. When plotting the intensity of peaks in the methyl region of Aβ, the low concentration 0.2 mg/ml sample showed the characteristic lag phase in which peak intensity was maintained for days, an elongation period marked by disappearance of detectable monomer from solution and, finally, full fibril formation and a complete loss of NMR signal after 12 days (Fig 1B). Interestingly, the higher concentration samples of 0.8 and 4.3 mg/ml initially disappeared more rapidly than the low concentration sample, but the signal was maintained long after the low concentration sample had fibrillized (Fig 1B). These signals persist at about 50% even a year later. Thus, an increasingly larger population of Aβ40 remains in solution as the peptide concentration is increased.

Interestingly, these high concentration oligomers can coexist with fibrils. As mathematical theory predicts, the high concentration environment favors formation of soluble oligomers ⁷, but does not preclude formation of sparse fibrils, which are visible by electron microscopy (Fig 1C). This is especially intriguing because the few fibrils present do not seed the soluble oligomers into joining the fibril fold—the energetic forces within the oligomer make it more favorable to remain in its existing conformation than to rearrange to join the energy-minimized amyloid conformation.

Characteristics of the stable oligomer

We employed SAXS to compare the particle conformations of the species formed by Aβ at different concentrations. SAXS plots of the log scattering intensity versus log magnitude of the scattering vector, Log(I(Q)) vs. Log(Q), show qualitative differences between measurements at various peptide concentrations (Fig 2A). At lower concentrations (0.2, 0.5, and 1 mg/ml), the Aβ curves display an approximately constant decrease in scattering intensity with Q, and appear to follow a roughly power law relationship, I(Q)∝Q^-n, where n > 0 (Fig 2A). The profile from the higher concentration sample (2.0 mg/ml) is clearly different. For example, the curve bends concave downward for Q < 0.1Å. These differences in how scattering intensity diminished with Q suggest that Aβ conformation is sensitive to peptide concentration. By processing the data into a Kratky plot (I(Q)*Q² vs Q), information can be obtained about the compactness of the protein. Compact proteins have ‘bell-shaped’ curves in which I(Q)*Q² increases with Q until it reaches a maximum, and then converges toward the Q-axis at large Q. Extended proteins have I(Q)*Q² that increases and then may plateau, or continue to increase at large Q. The Kratky plots show that the Aβ in the lower concentration samples (0.2, 0.5, and 1 mg/ml) are in extended conformations and are not well folded (A Gaussian coil polymer should have I(Q) ∝ Q^-2 asymptotic behavior and plateau in a Kratky plot.) In contrast, the 2 mg/ml Aβ sample reaches a maximum and then decreases, indicating the peptide oligomers are compact. The compactness of oligomers formed at higher concentrations is in good agreement with the kinetics experiments, considering that sample preparation differences such as agitation, incubation period, and buffer conditions can influence protein conformation. Interestingly, the curve remains above the Q-axis, suggesting there is a certain degree of flexibility in the Aβ oligomers at this concentration. Thus, the oligomers formed at the higher concentration are more compact and well folded than the unfolded species found at low Aβ concentration.

Figure 2.

Small-angle X-ray scattering (SAXS) shows that the Aβ peptide forms compact oligomers at high concentrations. A. Log-log SAXS plots show the curve from the high concentration 2 mg/ml Aβ sample has a different shape compared with low concentration (0.2, 0.5, and 1 mg/ml) curves. The spectra are offset vertically for clarity. B. Kratky plots, I(Q)*Q² vs. Q, of the SAXS spectra highlight that the higher concentration sample (2.0 mg/ml) contains compact oligomers, demonstrated by a characteristic peak with lower signal at high Q, while the lower concentration samples (0.2, 0.5, and 1 mg/ml) are in a more extended conformation, characterized by increased signal at high Q. The measured scattering intensities, I(Q), are in arbitrary units (AU), and Q is in inverse Ångström (Å^-1). The 0.2 mg/ml Aβ sample is at the lower concentration limit for SAXS and, therefore, exhibits some noisiness.

For the higher concentrations of Aβ, there is a small but gradual increase in the ThT signal at the early time points, but the exponential increase that indicates fibril formation occurs days later (Fig 1A). We thought this slight increase in fluorescence might indicate formation of oligomers that had some low level of reactivity with ThT. To examine this oligomer more closely, we used ANS as a probe to see if the peptides adopted molten globule conformations characteristic of disordered oligomers. ANS binds to nonpolar (hydrophobic) sites of proteins, but its fluorescence is quenched by water, so it is only detected when binding in a pocket that occludes water ³⁵. While ANS was able to detect hydrophobic pockets in the control protein BSA, it did not detect any molten globule character during Aβ oligomer formation (Fig 3A). Thus, the oligomer being formed is configured in a way that does not expose a hydrophobic pocket, suggesting that it is in a compact arrangement with tight contacts.

Figure 3.

Size analysis of the stable oligomer. A. The fluorescent dye ANS was used to probe for molten globule structure. During the timecourse of oligomer formation, the 1.0 mg/ml Aβ sample did not show any exposed hydrophobic pockets indicative of molten globules. B. A 1 mM Aβ solution was evaluated by SEC, either prepared immediately before injection or first aged for two weeks. The fresh sample (blue) showed immediate formation of a small oligomer, while the aged sample (red) acquired a population of larger conformers. Because DMSO was detected at the same molecular weight as the Aβ monomer, it is impossible to resolve the monomer accurately. C. An Aβ oligomer solution was prepared at 1 mg/ml, aged for two weeks, and characterized using AUC. Sedimentation velocity at 45,000 rpm and subsequent calculation of the continuous c(S) distribution showed a distribution of species, with the majority being a dimer. Average S = 3.544.

We used SEC to characterize the full distribution sizes of oligomers generated at high concentration at short and long times. Using a Superdex 75 column, we analyzed a 1 mM Aβ sample that was either prepared immediately before use or after being aged for two weeks. The size distribution of oligomers evolves as the sample ages (Fig 3B). When the sample was freshly prepared (blue), the peptides immediately populated low molecular weight (MW) oligomer states, roughly averaging around an apparent MW corresponding to a trimer to tetramer. Given that the peptide is only loosely folded and the MW standards used were well-folded globular proteins, it is likely that the apparent MW of a trimer or tetramer is an upper limit. After aging two weeks (red), the intensity of the signals decreased overall, and the peptides adopted a larger range of sizes, with some portion still maintaining the low-MW oligomer, but with the majority maturing into high-MW species (Fig 3B). We can conclude that the Aβ peptides almost immediately form small oligomers, and with time, they become much larger, possibly forming some fibrils.

We also used analytical ultracentrifugation (AUC) to evaluate the size of the stable oligomeric complex in a 1 mg/ml sample by sedimentation velocity at 45,000 rpm (Fig 3C). Calculating the continuous c(S) distribution showed that there were multiple species present, with the majority being 8.16 kD close to the value expected for a dimer (Fig 3C). Thus, although larger oligomers are present, the dimer seems to be the primary and minimal functional unit of the long-lived Aβ oligomer. Because SEC is a lower resolution method, the AUC estimate is likely the more accurate description of oligomer size.

Structural insights into the oligomer by NMR

We measured the 2D-NMR spectra of the long-lived oligomer by solution NMR and made sequential assignments in the H^N-H^α section of 2D NOESY (mixing time 150 ms) for the 1 mM Aβ oligomer sample (Fig 4A). We also used variable gradient diffusion experiments to evaluate the effective size of the oligomers. Effective diffusion coefficients of the integrated intensity over the full chain indicated that the chains were free to diffuse as a dimer or trimer over this temperature range(Fig. S4).

Figure 4.

2D-NMR spectra of the long-lived oligomer by NMR. A. Sequential assignments in the H^N-H^α section of 2D NOESY (mixing time 150 ms) for the 1 mM Aβ oligomer sample at ¹H frequency of 900 MHz at 298K in 10 mM NaPi, pH 7.2, 95/5v% H₂O/D₂O, 21v% DMSO. The spectrum was recorded 4 days after sample preparation. The assignments shown in green indicate intra H^N-H^α NOE, while those in red show sequential NOE between H^N and the preceding H^α. B. Using ¹³C,¹⁵N isotopically labeled Aβ peptide, the 2D ¹⁵N- HSQC spectra of the 1 mM Aβ oligomer sample has distinct and clearly resolvable peaks for each amino acid.

In order to gain greater insight into the structure of the oligomer, we used uniformly labeled ¹³C,¹⁵N Aβ to obtain a ¹⁵N- HSQC of 1 mM Aβ (Fig 4B). The methyl groups are fully accounted for when viewed in detail (Fig S1). The oligomer gave clear spectra at 37°C, suggesting a well-ordered conformation, but the spectra become even stronger at lower temperatures, suggesting the peptide complex becomes more well defined at low temperatures (Fig S2).

The backbone assignments show transfer of nuclear spin polarization (known as nuclear Overhauser effect [NOE]) for intra H^N-H^α NOE (green) and NOE between H^N and the preceding H^α (red). The fact that most sequential H^N-H^α NOEs are significantly stronger than the intra ones illustrates that the Aβ peptide takes an extended conformation. Unlike the data for an SDS-solubilized oligomer, no long-range mainchain–mainchain NOEs were observed. On the other hand, a number of medium- to long-range interactions were observed between sidechains. The lack of long-range backbone–backbone NOEs suggests that the beta-strands are parallel, rather than antiparallel as in the SDS-solubilized structure. However, it was not possible to compute a unique structure for the dimer due to the paucity of long-range NOEs. We therefore suggest that the peptides have a relatively fixed secondary structure but a fluctuating tertiary and quaternary structure.

It is instructive to compare the arrangement of the secondary structure of these water-soluble oligomers with the fibrils formed from Aβ40 ^{9, 16}, and the toxic Aβ42 oligomer form, as determined by solids NMR ²². The phi/psi values predicted from TALOS are often disordered for the first 10 residues; therefore, we compare the more well-structured region from residues 10-40. In all preparations, a beta conformation is observed from approximately 11-21, 30-33, and 34-37 (Fig. 4). There are, however, significant differences in the inter-strands region between the various structures. Our soluble oligomers differ primarily in having slightly longer predicted beta-strands. Thus, the beta-region of the soluble oligomer would appear to be largely pre-organized to form fibrils. It is possible, however, that the twist of the strands prevents further edge-to-edge hydrogen-bonding, which is necessary to propagate fibril formation. Thus, the dimers appear to interact in a more flexible manner leading to a fluctuating structure that does not progress to form fibrils.

Discussion

We have demonstrated that very high concentrations of Aβ encourage the peptide to forego fibril formation and instead associate to a soluble oligomer consisting of a stable dimer in equilibrium with higher-order aggregates( Fig. 6). We showed that above a certain threshold, the lag phase of Aβ fibrillization by ThT increases, and this delay of fibril formation was confirmed by solution NMR. These findings are consistent with the results of Jackson and coworkers, who have recently demonstrated the presence of pH-sensitive oligomers, which compete with fibril formation during aggregation of glucagon-like peptide ²⁶. The high concentration assemblies of Aβ40 were more compact than the corresponding monomer as measured by SAXS and were determined to consist primarily of dimers by AUC. However, size exclusion chromatography showed that at longer times this dimer is in equilibrium with higher order oligomers, which might impart the stability at high peptide concentrations relative to the monomer and fibril. Solution NMR experiments showed relatively sharp peaks and an apparent diffusion coefficient similar to that of a dimer. The most straightforward explanation of these data is that the peptide forms a stable dimer, which is loosely associated into higher order oligomers. Most likely, the dimers average between many different orientations within the oligomer and can exchange between oligomers, giving rise to relatively sharp peaks despite the large apparent molecular weight seen by SEC.

Fig. 6.

Schematic diagram of the monomer-oligomer-fibril reaction and equilibrium. The energy levels of the intermediate depend more strongly on concentration than the fibril state.

The extreme kinetic stability of this compact oligomer, and apparent reversibility upon heating and cooling, suggests that the oligomer represents an ensemble that is energetically more favorable than the cross-beta conformation in amyloid once the concentration of Aβ exceeds a specific threshold. While this fits with the mathematical predictions by Powers and Powers ⁷, it is still surprising given the conventional wisdom that increased concentration leads to faster fibrillization. The thermodynamic stability of the oligomer would be high order with respect to monomer concentration, while the addition of a monomer to a fibril is first order in the monomer concentration. Thus, at a sufficiently high peptide concentration, soluble oligomers and fibrils can coexist, as seen in our experiments. It is also possible that the observed stability is kinetic in origin: at high concentrations, the Aβ40 dimers might encounter each other so rapidly that formation of an off-pathway oligomer is more favorable than the structural rearrangement required to form an obligate oligomer that seeds fibril formation.

In summary, we have identified a highly stable oligomer that persists for at least a year. Even in the presence of usually highly replicative amyloid fibrils, the oligomer is stable enough to resist conversion into the fibril fold. Therefore, this may be an attractive target for AD treatment: using therapeutic molecules to encourage Aβ to adopt this stable oligomer rather than proceeding down the amyloid cascade. This approach may prevent the generation of transient toxic oligomers, the formation of nucleation sites for fibrillization, and the recruitment of naïve Aβ peptides to existing fibrils and plaques within the AD brain.

Figure 5.

Comparison of phi/psi from the current solution sample with those from solid-state NMR studies from ²² (labeled “Ishii”) , 2LMN , 2LMP, and 2M4J. All dihedral angles were based on TALOS prediction from chemical shifts or from structures.