Fibrillar Oligomers Nucleate the Oligomerization of Monomeric Amyloid β but Do Not Seed Fibril Formation*

Soluble amyloid oligomers are potent neurotoxins that are involved in a wide range of human degenerative diseases, including Alzheimer disease. In Alzheimer disease, amyloid β (Aβ) oligomers bind to neuronal synapses, inhibit long term potentiation, and induce cell death. Recent evidence indicates that several immunologically distinct structural variants exist as follows: prefibrillar oligomers (PFOs), fibrillar oligomers (FOs), and annular protofibrils. Despite widespread interest, amyloid oligomers are poorly characterized in terms of structural differences and pathological significance. FOs are immunologically related to fibrils because they react with OC, a conformation-dependent, fibril-specific antibody and do not react with antibodies specific for other types of oligomers. However, fibrillar oligomers are much smaller than fibrils. FOs are soluble at 100,000 × g, rich in β-sheet structures, but yet bind weakly to thioflavin T. EPR spectroscopy indicates that FOs display significantly more spin-spin interaction at multiple labeled sites than PFOs and are more structurally similar to fibrils. Atomic force microscopy indicates that FOs are approximately one-half to one-third the height of mature fibrils. We found that Aβ FOs do not seed the formation of thioflavin T-positive fibrils from Aβ monomers but instead seed the formation of FOs from Aβ monomers that are positive for the OC anti-fibril antibody. These results indicate that the lattice of FOs is distinct from the fibril lattice even though the polypeptide chains are organized in an immunologically identical conformation. The FOs resulting from seeded reactions have the same dimensions and morphology as the initial seeds, suggesting that the seeds replicate by growing to a limiting size and then splitting, indicating that their lattice is less stable than fibrils. We suggest that FOs may represent small pieces of single fibril protofilament and that the addition of monomers to the ends of FOs is kinetically more favorable than the assembly of the oligomers into fibrils via sheet stacking interaction. These studies provide novel structural insight into the relationship between fibrils and FOs and suggest that the increased toxicity of FOs may be due to their ability to replicate and the exposure of hydrophobic sheet surfaces that are otherwise obscured by sheet-sheet interactions between protofilaments in a fibril.

The accumulation of aggregated amyloid proteins is a characteristic hallmark of a wide range of human degenerative diseases, including Alzheimer (AD), 2 type II diabetes, Huntington, Parkinson, and spongiform encephalopathy (1,2). Although familial mutations that result in increased production of A␤42 support a causal role of A␤ peptide, the mechanism of A␤ pathogenesis remains unknown and controversial. A␤ plaques composed of A␤ fibrils are evident in both patients with AD and a significant number of healthy individuals who are cognitively normal (3). Emerging evidence implicate soluble oligomers formed during protein aggregation are the primary toxic species in amyloid pathogenesis (4). Several studies have shown that A␤ oligomers are toxic in cell culture and transgenic animal models (5)(6)(7)(8)(9)(10). A␤ oligomers are found in AD brain, where they have a localization pattern that is distinct from diffuse and thioflavin-positive plaques (5). The increased level of soluble oligomers also correlates better than plaque load with cellular dysfunction and the severity of cognitive impairment in AD patients and transgenic animal models (5,7,11). The observation that amyloid oligomers derived from protein sequences that are not related to a human disease are toxic to cell culture models (12) suggests an inherent and generic toxicity of amyloid oligomers.
Although these reports of toxic oligomers are in agreement in terms of solubility, their size, morphology, and immunoreactivity varies indicating that amyloid oligomers are structurally diverse (2). A␤ oligomers have been prepared by several different protocols and variously referred to as A␤-derived diffusible ligands, globulomers, protofibrils, and amyloid spheroids (6,(13)(14)(15). Oligomers that correlate with cognitive dysfunction in transgenic animals have been reported to be 56 kDa (A␤*56) as determined by SDS-PAGE, whereas oligomers purified from cell culture and from human AD brains have been reported to migrate at the position of dimers (9,16,17). Synthetic A␤ oligomers between 50 and 100 kDa in size are also reported as toxic and highly correlative to synaptic loss in cell culture models (11,18). A similar spectrum of soluble oligomers has been observed for other amyloidogenic proteins, including islet amyloid polypeptide, ␣-synuclein, prion, insulin, and lysozyme (19 -23). It is unknown whether these oligomeric species represent the same or distinct structural variants. Discriminating among the different structures of the aggregated * This work was supported, in whole or in part, by National Institutes of Health species is the first step to advance our understanding of the aggregation mechanism of amyloid proteins and to clarify the role of oligomers in amyloid pathogenesis.
In the absence of high resolution x-ray crystallographic structural data, conformation-specific antibodies provide a means of distinguishing and classifying the different types of amyloid aggregates by their underlying structures. We have previously reported three antibodies, A11, OC, and ␣-annular protofibrils, that recognize mutually exclusive generic epitopes specifically associated with prefibrillar oligomers (PFOs), fibrils, and annular protofibrils formed of different amyloid proteins, respectively (5,24,25). Interestingly, the OC antibody also recognizes small and soluble oligomers that share the same epitope as fibrils (24). The fact that FOs and fibrils share the same epitope suggests that FOs are structurally similar to fibrils and distinct from PFOs. Because fibril formation is known be nucleation-dependent, the simplest interpretation is that FOs represent fibril nuclei or small pieces of fibril filaments organized in the same structural lattice as in fibrils.
In this study, we examined the properties of A␤ fibrillar oligomers and their relationship to A␤ fibrils in greater detail. Although FOs are immunologically related to fibrils, their morphology, size, and tinctorial properties are distinct from fibrils. Surprisingly, we found that A␤ FOs do not seed the formation of ThT-positive fibrils from A␤ monomers. Instead, FOs nucleate the formation of soluble FOs that are ThT-negative, positive for the anti-fibril antibody, OC, and have the same size and morphology as the FOs seeds. Taken together, these data suggest that FOs may represent small pieces of soluble single fibril protofilaments that are capable of replication by the addition of monomers to the ends of FOs and subsequent splitting. This suggests that the replication of FOs may be a key determinant of the enhanced pathogenic activity of amyloid oligomers.

EXPERIMENTAL PROCEDURES
A␤ Preparations-For preparation of monomers, synthetic A␤40 (0.3 mg) was solubilized in 150 l of hexafluoro-2-propanol (HFIP), evaporated to dryness under a stream of N 2 , diluted in H 2 O, and used immediately. FOs were prepared by first dissolving A␤ in HFIP at a concentration of 420 M for 25 min at room temperature. Then the peptide solution was diluted into double distilled H 2 O (0.02% sodium azide), 70 M, and the pH was adjusted to 7.4 with NaOH. The sample was stirred at 500 rpm with a Teflon-coated micro stir bar for 2-3 days at room temperature with a cap that has holes. PFOs were prepared as described previously (5). The preparation was similar to that of FOs except the pH was adjusted to 2.5 with HCl. Fibrils were prepared by dissolving A␤ (2 mM) in 100 mM NaOH, followed by bath sonication for 30 s, and then stirring in 10 mM HEPES, 10 mM NaCl, 0.02% sodium azide, pH 7.4, at 500 rpm for 9 -10 days. Solubility was assayed by centrifuging A␤ solutions at 100,000 ϫ g for 1 h at room temperature. The pellets were collected, washed three times in double distilled H 2 O, and resuspended in desired buffers.
Dot Blots-Dot blot assays were performed to detect A␤ PFOs, FOs, and fibrils as described previously (5). In the case of the FO seeding experiment, the fibril-specific antibody LOC was used. The LOC antibody was generated against the islet amyloid polypeptide fibrils, and like OC, it recognizes a variety of fibrils and FOs formed from different amyloidogenic proteins and peptides, including A␤, but it does not detect A␤ monomers or PFOs (24). In using the LOC antibody, we further ensured that only A␤ fibrils and FOs, and not A␤ monomers, will be detected. Briefly, 1-l spots containing 0.3 g of A␤40 were applied to a nitrocellulose membrane and blocked in 10% nonfat dry milk in Tris-buffered saline (TBS-T) containing 0.01% Tween 20 at room temperature for 1 h. After three 5-min TBS-T washes, the membranes were probed with conformation-specific antibodies (A11, 1:2000; OC, 1:10,000; LOC, 1:2000) overnight at 4°C. The membranes were then incubated with anti-rabbit IgG conjugated with horseradish peroxidase (1:10,000, Promega) at room temperature for 1 h. The blots were developed with ECL chemiluminescence kit from Amersham Biosciences.
Western Blot-Samples containing 4 g of A␤40 were dissolved in SDS treatment buffer, boiled for 5 min, and electrophoresed on 10 -20% Tris-HCl (Bio-Rad) gels. Proteins were electrophoretically transferred onto nitrocellulose membranes and developed with conformation-specific antibodies as described above.
ThT Fluorescence Assay-ThT fluorescence was used to monitor A␤ fibril formation, as described previously (26). Samples of A␤ conformers were mixed with ThT solution (3 M in 10 mM Na 2 HPO 4 , pH 7.4), and fluorescence was measured in a Gemini XPS plate reader (Molecular Devices, Sunnyvale, CA). For seeding experiments, reactions were performed in 96-well plates, and the ThT fluorescence signals were read at 10-min intervals. All experiments were performed in triplicate.
Electron Microscopy-One-l aliquots of A␤ monomers, PFOs, FOs, and fibrils were adsorbed onto 200-mesh Formvar/ carbon-coated nickel grids until dry. The grids were then washed with water, stained with 2% uranyl acetate, and washed again. The grids were allowed to dry between all steps and were viewed in a Phillips CM 12 microscope operated at 65 kV.
Atomic Force Microscopy-AFM analysis was carried out using a Nanoscope III multimode instrument (Veeco Instruments, Santa Barbara, CA). 1.5 l of sample solutions were deposited on freshly cleaved mica coated with poly-L-lysine (0.01% w/v) for 1-2 min. Then the sample was rinsed with double distilled H 2 O and dried under a stream of N 2 gas. Imaging was done in tapping mode with a silicon cantilever at a scan frequency of 1 Hz and an oscillation frequency of 165 kHz. The heights of individual fibrils were measured perpendicular to the fibrillar axis (in case of fibrils) using the profile tool in SPIP software (Image Metrology) and using scan sizes of less than 3 m 2 . Height analysis was calculated using Voss volume software (27).
FTIR-FTIR spectra were measured with a Bruker Equinox 55 FTIR instrument (Bruker Optics, Billerica, MA) equipped with a deuterated triglycine sulfate detector. Aqueous solutions of either fibrils or FOs of A␤40 were deposited on the calcium fluoride glass and dried under nitrogen. 512 scans at 2 cm Ϫ1 resolution were collected for each sample under constant purging with nitrogen and corrected for water vapor, and background spectra were subtracted.
Preparation of Spin-labeled A␤ Peptides-Spin-labeled A␤40 derivatives were generated through labeling of single cysteine mutants. Single cysteine mutants were synthesized, purified, and verified by mass spectrometry as described previously (28). For labeling, lyophilized A␤40 cysteine mutants were initially dissolved in 100% DMSO and then diluted to 50% with Milli-Q water. 3-Fold molar excess of (1-oxy-2,2,5,5-tetramethypyrroline-3-methyl)-methanethiosulfonate spin label was used to label the peptide for 1 h at room temperature. Hydrochloric acid was then added to a final concentration of 1 mM, and excess spin label was removed using a cation exchange column (Toyopearl resin, Tosoh Corp.). The peptides were washed with a minimum of 10 column volumes of 1% acetic acid or until no more free spin label was detectable by EPR spectroscopy. The spin-labeled peptides were eluted from the column using buffer containing 6 M guanidine HCl in 100 mM Tris, pH 7.4. Subsequently, C-18 Macro SpinColums (Harvard Apparatus) were used for desalting, washing, and eluting of the spin-labeled peptides in 100% HFIP. Finally, the concentration was determined using absorbance at 280 nm and an extinction coefficient of 1280 for the single tyrosine. Samples were aliquoted and lyophilized for later use.
EPR Analysis of A␤ Oligomers-To prepare fibrillar oligomers for EPR analysis, lyophilized spin-labeled peptide were resuspended in HFIP (stock concentration of 1 mM peptide) for 30 min at room temperature. This stock solution was then diluted 1 to 4 into a 10% HFIP solution, giving a final peptide concentration of ϳ250 M and an HFIP concentration of greater than 20%. HFIP was then removed by evaporation with a stream of N 2 gas for about 10 min. Small aliquots (1-2 l) were saved for dot blot analysis with OC and A11 antibodies. Oligomers prepared in this way were positive to the OC antibodies but not to the A11 antibodies. Prefibrillar oligomers were made and assayed by dot blot as described previously (5). EPR spectra were recorded at room temperature with a Bruker EMX X-band EPR spectrometer fitted with an ER4119HS resonator. Some spectra contained up to 2% spectral components arising from monomeric unfolded peptide. These components were subtracted, and the resulting spectra were normalized to same number of spins and base-line corrected. All spectra are shown with a magnetic field scan width of 150 G.

Kinetics of Soluble Oligomer Formation-We have previously
reported that the fibril-specific, conformation-dependent antibody, OC, also recognizes 100,000 ϫ g soluble oligomers that range in size from dimer to greater than 250 kDa on Western blots (24). To characterize the relationship of these oligomers to fibrils, we investigated the kinetics of FO in comparison with fibril formation. FO formation was monitored by OC immunoreactivity, whereas fibril formation was measured by centrifugation at 100,000 ϫ g (Fig. 1, A and B). OC immunoreactivity of A␤ monomer was minimal at the initial time point and increased after 24 and 48 h (Fig. 1A). No A11 immunoreactivity was detected under these conditions, indicating that PFOs are not formed under these conditions. 6E10 immunoreactivity was uniform over the time course, indicating that equal amounts of peptide were spotted and demonstrating the conformation-specific nature of the OC epitope. OC immunoreactive A␤42 FOs are completely soluble at 100,000 ϫ g after 48 h of incubation, whereas most of the OC immunoreactive material sediments after 72 h of incubation (Fig. 1B). These data demonstrate that soluble FOs appear in solution prior to the formation of sedimentable fibrils and that FOs eventually form fibrils.
Fibrillar Oligomers Are Rich in ␤-Sheet Structure but Weakly Bind ThT-We further analyzed the secondary structure of A␤ FOs using FTIR spectroscopy. The FTIR spectrum of A␤42 FOs showed the presence of a high content of ␤-sheet structure with prominent bands at 1629 and 1698 cm Ϫ1 (Fig. 2). Additionally, FOs exhibited a broader shoulder at 1640 -1670 cm Ϫ1 indicative of a higher content of random coil, ␣-helices, and ␤-turns. The spectrum of FOs was very similar to the spectrum of A␤42 fibrils except the intensity of the shoulder at 1640 -1670 cm Ϫ1 was lower in fibrils indicating that they contain a higher percentage of ␤-sheets. The spectral similarity between the spectra  of FOs and fibrils is consistent with their common immunoreactivity with the fibril-specific conformation-dependent antibody, OC, which indicates that they display the same conformational epitope.
To further compare the properties of FOs and fibrils, we examined their ability to bind to ThT. ThT is a fluorophore that selectively exhibits fluorescence at 480 nm upon binding specifically to amyloid fibrils in vitro and in vivo and not to amyloid precursor protein, monomers, or other ␤-sheet-rich amyloid aggregates (29). We freshly prepared A␤42 and A␤40 monomers, FOs, and fibrils and measured the dependence of ThT fluorescence on the concentration of these A␤ aggregates. A␤40 monomers exhibited base-line ThT fluorescence as expected (Fig. 3). In comparison, both A␤40 and A␤42 FOs weakly enhance the fluorescence of ThT. The extent of fluorescence enhancement by FOs is similar to that of PFO, which has been reported previously (30). However, the fluorescence enhancement of fibrils is 33-fold higher than FOs (Fig. 3). This difference suggests that there is a structural difference between fibrils and FOs that is not apparent by OC immunoreactivity.
Fibrillar Oligomers Are Structurally Related to Fibrils and Distinct from PFOs-To compare the structural features of the different oligomer preparations, we generated spin-labeled A␤ oligomers that were either positive to the OC (FOs) or A11 antibody (PFOs) and recorded their EPR spectra (Fig. 4). We have previously shown that spin-labeled A␤ peptides retain the ability to form amyloid fibrils and co-assemble normally with unlabeled A␤ (28). As illustrated with the example of A␤ labeled at position 14 (14R1), FOs and PFOs give rise to distinctly different spectra. The most striking difference is the reduced signal amplitude of OC-positive FOs, which is caused by significant line broadening that extends beyond the typical ϳ70 G width of a nitroxide spectrum (Fig. 4B). This broadening is a clear indicator of strong spin-spin interaction that is present in the OC-positive oligomers. Although some spin-spin interaction is also visible in the A11-positive PFOs, it is clearly much smaller and only has minor spectral contributions. These results indicate that the spin labels in PFOs are farther apart than in fibrils and FOs. To further demonstrate the spin-spin interaction in the OC-positive FOs, we performed dilution experiments containing only 10% R1-labeled proteins. Under these conditions, the effects of spin-spin interaction should be strongly alleviated, and in fact, a pronounced increase in signal intensity and loss of the broadening can be observed (Fig. 4C).
Overall, the strong spin-spin interaction makes the EPR spectra of the OC-positive oligomers more similar to those of the fibrils, whose parallel, in-register structure brings neighboring labeled sites into close contact (Fig. 4A). Despite these similarities, however, analysis of the EPR spectra for fibrils and FOs reveals some differences also. The EPR spectrum for the fibril contains a significant spectral component of spin exchange narrowing (shown after spectral subtraction in Fig. 4D). This component indicates the simultaneous contact (orbital overlap) of multiple spin labels in agreement with a parallel in-register structure formed by multiple strands (31,32). No such component can readily be detected in the case of the OC-positive FOs. Moreover, comparison of the EPR spectra in Fig. 4C reveals that there is a significant distribution of distances between labeled side chains from close to contacts to distances that are in the 10 -20 Å range. Thus, although the FOs are likely structurally related to the fibrils, there appears to be significantly more heterogeneity in spin-spin interaction. This notion is further supported by spectral analysis of several other labeled sites that display the same spectral features (supplemental Fig. 1). In general, the EPR spectra from all sites indicate that the OC-positive FOs (and fibrils) have significantly more spin-spin interaction than the A11-positive PFOs. The only site that gives rise to relatively little spin-spin interaction for all of the forms is the carboxyl-terminal residue at position 40.  A shows the EPR spectrum of the fibrils, which was reported previously and is reproduced here for comparison. The EPR spectra for FOs (B, solid line) are of much lower amplitude than those of PFOs (B, dashed line). The reduction in amplitude coincides with significant spectral broadening beyond 70 G indicating the presence of strong spin-spin interaction. The presence of spin-spin interaction is illustrated in C, which overlays the spectrum of FOs labeled at 100% (solid line) with the spectrum of FOs generated from 10% of spin-labeled peptide and 90% of wild-type peptide (dashed line). In contrast to the FOs, the spectra of the fibrils contain a clear component of spin exchange, which is characterized by single line EPR spectra (D) (32). The spectral component was obtained by subtraction of the hyperfine structure from the spectrum in A. Similar spectral subtractions did not yield any evidence for single line, exchange narrowed EPR spectra for FOs. All spectra were obtained at the X-band, normalized to the same number of spins, and are shown at a scan width of 150 G.
Fibrillar Oligomers Are One-half to One-third the Height of Fibrils-To gain insight into the morphological relationship between PFOs, FOs, and fibrils, we examined them by atomic force microscopy. PFOs appeared as large globular aggregates that are 20 -40 nm in diameter, whereas FOs are smaller aggregates that are 5-15 nm in diameter (Fig. 5, A and B). Fibrils, on the other hand, were straight, 10 -20 nm in width, 500 -900 nm long, and consisted of multiple filaments (Fig. 5C). Fibrils also exhibit morphological diversity with respect to their width, length, and degree of twisting, consistent with the well known structural polymorphism of fibrils (33). Height analysis reveals that the height of FOs is distributed within a narrow range (1.13 Ϯ 0.08 nm) that is consistent with the presence of a single sandwich of two ␤-sheets (Fig. 5D). Based on their dimensions in AFM, FOs are calculated to contain 3-10 A␤ monomers in agreement with their size distribution on Western blots (supplemental Fig. 2). For the calculation, the volume of the oligomer was obtained from the AFM and EM data, and the volume of each monomer was estimated assuming peptide partial specific volume of 0.73 cm 3 /g (27). AFM analysis revealed the heights of A␤42 fibrils are in a broad range of 2-5 nm (Fig. 5D). The fact that FOs are one-half to one-third the height of fibrils suggests that they may represent small segments of the protofilaments that make up mature fibrils.

Fibrillar Oligomers Nucleate the Oligomerization of Monomeric A␤ and Do Not Seed Fibrils-Because
FOs are small and soluble oligomers that share secondary structure and immunoreactivity with fibrils, it is possible that they may represent fibril nuclei or seeds. Alternatively, FOs may represent a unique type of A␤ oligomers that share conformational similarity to fibrils. To address this question, we analyzed the seeding properties of FOs. The assembly of amyloid fibrils is a nucleation-dependent process, and assembly reactions display a characteristic lag phase prior to initiation of fibril formation that reflects the time required for the spontaneous formation of fibrils (34). This lag phase is reduced or eliminated by seeding the assembly reaction with preformed fibrils. Mature A␤40 fibrils that have been sheared by sonication were used to seed monomeric A␤40 as a positive control, and A␤40 monomers without any seeds served as a control. In the absence of preformed seeds, monomeric A␤40 assembled into ThT-positive fibrils with a lag phase of 2.79 Ϯ 0.33 h (Fig. 6A). The presence of 5 and 10% of fibrils reduced this lag phase to 2.64 Ϯ 0.64 and 1.14 Ϯ 0.13 h, respectively (supplemental Table 1). In contrast, the addition of 5 and 10% of FOs slightly prolonged the lag phase of fibril assembly to 3.03 Ϯ 0.29 and 3.01 Ϯ 0.34 h (Fig. 6B). This indicates that FOs do not seed fibril formation and are not fibril nuclei.
Because FOs do not seed fibrils, we then examined whether FOs seed the formation of FOs using the fibril-specific and conformation-specific antibody LOC as a read-out of FO formation (24). Assembly reactions containing 2% of A␤40 FOs and fibrils were assayed at different time points for fibril-specific immunoreactivity (Fig. 7A). A␤40 monomers with no seeds were used as controls. For A␤40 monomers seeded with 2% of FOs, significant LOC immunoreactivity was observed as early as 0.5 h and increased up to 2 h after the initiation of aggregation (Fig. 7A). In contrast, unseeded assembly reactions and solutions seeded with A␤ fibrils did not develop LOC immunoreactivity until 1.5-2 h (Fig. 7A). The increase in fibril immunoreactivity is due to the conversion of monomer into FOs, because the added seeds are below the limit of detection at this dilution (Fig. 7B). In contrast, A␤40 monomers incubated without seeding exhibited slower aggregation kinetics. LOC immunoreactivity was minimal at 1.5-h time point and became strong only at 7 h after initiation of aggregation (Fig. 7, A and B).
Fibril immunoreactivity could be indicative of the formation of either small FOs or fibrils; however, the lag time for fibril formation as judged by ThT fluorescence under the same conditions is 2.6 h, suggesting that fibrillar oligomers had formed rather than mature fibrils. These results suggest that FOs seed the formation of FOs rather than fibrils, and the kinetics of FO formation are faster than fibril formation in seeded reactions. To confirm this, TEM at the 1-h time points revealed the appearance of small FOs (5-15 nm in diameter) that are not observed at time 0 in A␤40 monomers seeded with FOs (Fig. 8,  A and B). After 2 h of aggregation, the number of oligomers increased. We have observed that the morphology and size (6 -15 nm) are faithfully propagated from the parental FO seeds onto the resulting daughter FOs (Fig. 8 and Fig. 9). No fibrils were observed in FO-seeded samples over the time frame of 0 -2 h (Fig. 8, A, B, and E). Consistent with the dot blot and ThT assays, TEM analysis revealed that the unseeded A␤40 sample did not contain oligomers over the same times of incubation (Fig. 8, A-D and G). The appearance of fibrils in fibril-seeded samples was evident at 2 h after initiation of aggregation and is correlated by TEM with the appearance of fibrils (Fig. 8, A, C,  and F). At longer incubation times, all reactions displayed the presence of mature fibrils in TEM images (Fig. 8, H-J). The presence of FOs in FO-seeded samples and the absence of observable fibrils provide further confirmation for the interpretation that FOs seed FO formation but not fibrils.

DISCUSSION
These results demonstrate that FOs represent a novel class of A␤ oligomers that are distinct from both PFOs and fibrils. The results show that A␤ FOs are soluble oligomeric species that are structurally and immunologically related to fibrils and are conformationally distinct from PFOs. FOs and PFOs are both soluble at 100,000 ϫ g, exhibit a high content of ␤-structure, are weakly thioflavin T-positive, and appear as spherical oligomers in EM and AFM images (5,24,30). However, FOs are smaller than PFOs by AFM analysis (Fig. 4) and are differentially recognized by A11 and OC antibodies, suggesting that they have a fundamental structural difference. Because both A11 and OC recognize generic epitopes that are displayed by many different amyloid sequences, these structural differences appear to represent fundamental classes of amyloid oligomer structures. These structural differences are readily apparent by EPR spectroscopy, where FOs display significantly stronger spin-spin interaction between adjacent spin labels, indicating that the labels are more closely spaced than PFOs ( Fig. 4 and supplemental Fig. 1). The lack of strong spin coupling from adjacent spin labels in PFOs suggests that the ␤-strands may be antiparallel or staggered or both. FOs share the same conformationspecific, generic epitope that is displayed in A␤ fibrils, which are known to be organized as parallel in-register ␤-sheets (28,(35)(36)(37). The EPR spectra of FOs and fibrils are very similar, but FOs lack evidence of spin exchange that arises from the molecular contact of multiple spin labels in linear arrays. Although FOs show some strong spin-spin interactions, the spin coupling is more heterogeneous than that observed in fibrils. Whether this heterogeneity is due to reduced coupling at the probes located at the ends of the small sheets or whether the FOs are polymorphic or structurally heterogene-FIGURE 6. A␤40 FOs do not serve as nuclei for fibril formation. A␤40 fibrils (A) and FOs (B) were sonicated for 5 min in a water bath and incubated at 25°C with 45 M A␤40 monomers (Mon). The addition of fibril seeds to monomers accelerated fibril formation (A). In contrast, when A␤40 monomers were seeded with FOs, the lag phase of fibril formation increased slightly in a seed percentage-dependent manner (B). Experiments were performed in triplicate. ThT data were normalized by subtracting ThT fluorescence values of fibril seeds and FO seeds incubated alone and buffer, respectively. AU, arbitrary units. A, A␤40 monomers seeded with 2 and 5% of FOs formed LOC-immunoreactive aggregates within 1.5 h. In contrast, A␤40 monomers in the absence of seeds developed LOC immunoreactivity after 1.5 h. Monomers seeded with fibrils developed LOC reactivity at 1.5 and 2 h. B, A␤40 FO and fibril seeds control. Immunoreactivity of seeds diluted to 2% is below the limit of LOC detection. C, quantification of LOC immunoreactivity of A␤40 monomers alone and monomers seeded with 2% FOs. The amount of LOC reactivity represents the amount of FOs or fibrils in each reaction. AU, arbitrary units. ous is not yet clear. A␤ fibrils are known to be polymorphic, and this variation arises from differences in the location of the bend in the ␤-hairpin (33,37), although it is not obvious that this type of polymorphism would account for the observed heterogeneity in spin coupling.
In contrast to fibrils, FOs are soluble at 100,000 ϫ g, are spherical in morphology, and bind ThT significantly less than fibrils. AFM analysis indicates that FOs are approximately one-third the heights of fibrils (1.13 nm for FOs and 3 nm for fibrils; Fig. 5). Although the absolute height estimates of proteins by AFM may be too low due to deformation of the protein by the AFM tip, the ratio of heights of FOs and fibrils should be an accurate reflection of their relative heights. In A␤40 fibrils, each peptide molecule is folded into a hairpin with ␤-strands (13-24 and 28 -40) separated by a loop, and these molecules are intermolecularly hydrogen-bonded to form a protofilament, with two or three protofilaments in a fibril, depending on their particular morphology (33). FOs are approximately one-half to one-third the height of fibrils by AFM image analysis. Taken together, these results suggest that FOs represent small segments of a protofilament (Fig.  10). Based on these results, we hypothesized that FOs may represent fibril nuclei, but surprisingly, we found that FOs are unable to nucleate fibril formation. Instead, FOs seed the formation of more FOs. The fact that FOs seed the replication of FOs but not fibrils is consistent with the suggestion that they have distinct growing ends or lattices. Because FO formation precedes fibril formation, this suggests that the formation of a protofilament lattice consisting of intermolecularly hydrogenbonded A␤ hairpins is kinetically more favorable than the formation of the fibril lattice, which contains two or more protofilaments that form sheet stacking interactions via their carboxyl-terminal sheet surfaces (Fig. 10). After 72 h of incubation, the majority of A␤ has assembled into insoluble fibrils, which indicates that fibrils are thermodynamically more favorable than fibrillar oligomers. Presumably, this increased stability results from the additional sheet stacking interactions of the hydrophobic surfaces of the carboxylterminal sheets of two or more adjacent protofilaments. The finding that FOs are weakly thioflavin-positive suggests that thioflavin binding may depend on the sheet stacking interface between the protofilaments.  FOs also nucleate the conformational conversion of monomer in a fashion distinct from that of fibrils. In fibril-seeded reactions, monomer conversion results in fibril elongation, which is monitored by an increase in thioflavin T fluorescence. If fibril nuclei contain intermolecular sheet stacking between 2 or more ␤-sheets, this suggests that the sheet stacking interaction between FOs that would be necessary to create a fibril nucleus is the rate-limiting step for fibril nucleation. However, FOs seed the conversion of A␤ monomer into FOs that are thioflavin-negative and are detected as an increase in OC immunoreactivity. Unlike fibrils, which appear to grow by addition of monomers, the FOs resulting from seeded reactions are not detectably larger than the FO seeds. This suggests that FOs grow to a size where their lattice becomes unstable and then split to produce additional growing ends. This implies that unlike A␤ fibrils, FOs are capable of replicating in a binary fashion that is similar to yeast prions.
The identification of A␤ FOs as small pieces of an individual protofilament suggests that FOs may be a generic oligomer class that is readily adapted by other amyloidogenic proteins, including islet amyloid polypeptide and expanded polyglutamine repeat protein peptides (24). Therefore, the ability of amyloid oligomers to propagate by seeding could also be a general property, intrinsic to other amyloidogenic proteins and other types of oligomers. Because amyloid oligomers have been reported to be more toxic than amyloid fibrils and they are widely believed to represent the primary pathological species in neurodegenerative disease (5,6,38), seeding could be one of the mechanisms by which amyloid formation spreads in various neurodegenerative diseases. For yeast prions, the propensity of the aggregates to fragment and generate new seeds is a key determinant of the ability of the prions to infect and propagate efficiently (39). By extension, the ability of FOs to fragment and generate new seeds may also be important for the enhanced pathogenic significance of oligomers as compared with fibrils in AD. The small size of FOs may also be important for toxicity as their ability to diffuse may be important for the spread of disease through the brain. The structure of FOs as pieces of amyloid protofilaments may also contribute significantly to their pathogenic activities. In fibrils, the hydrophobic carboxyl-terminal sheet surface from residues 28 -40 is shielded by sheet stacking interaction with the same surface from adjacent protofilaments, but in FOs, this hydrophobic surface is exposed.
This exposed hydrophobic surface may be able to interact with other hydrophobic surfaces, like cell membranes, and this interaction may explain the ability of amyloid oligomers to interact adventitiously with hydrophobic interfaces and permeabilize cell membranes (40 -42). FIGURE 10. Schematic diagram of amyloid aggregation pathways. Misfolded or random coil monomers can assemble along two divergent pathways into two distinct types of oligomeric intermediates as follows: prefibrillar oligomers and fibrillar oligomers. Although the structure of PFOs is unknown, we propose that FOs are pieces of single fibril protofilament. Fibrils, on the other hand, contain two or three protofilaments composed of A␤ organized in parallel in-register ␤-sheets. Only the end of the aggregates are shown, but they contain multiple intermolecularly hydrogen-bonded strands. Once FOs or fibrils have formed, monomer can add on to the ends, resulting in elongation of fibrils and replication of FOs that appear to reach a size at which they become unstable and break into smaller units. It is not yet clear whether FOs coalesce via stacking of the two carboxyl-terminal sheets or whether they dissociate into monomer prior to fibril formation.