Aβ42 Neurotoxicity Is Mediated by Ongoing Nucleated Polymerization Process Rather than by Discrete Aβ42 Species*

The identification of toxic Aβ species and/or the process of their formation is crucial for understanding the mechanism(s) of Aβ neurotoxicity in Alzheimer disease and also for the development of effective diagnostic and therapeutic interventions. To elucidate the structural basis of Aβ toxicity, we developed different procedures to isolate Aβ species of defined size and morphology distribution, and we investigated their toxicity in different cell lines and primary neurons. We observed that crude Aβ42 preparations, containing a monomeric and heterogeneous mixture of Aβ42 oligomers, were more toxic than purified monomeric, protofibrillar fractions, or fibrils. The toxicity of protofibrils was directly linked to their interactions with monomeric Aβ42 and strongly dependent on their ability to convert into amyloid fibrils. Subfractionation of protofibrils diminished their fibrillization and toxicity, whereas reintroduction of monomeric Aβ42 into purified protofibril fractions restored amyloid formation and enhanced their toxicity. Selective removal of monomeric Aβ42 from these preparations, using insulin-degrading enzyme, reversed the toxicity of Aβ42 protofibrils. Together, our findings demonstrate that Aβ42 toxicity is not linked to specific prefibrillar aggregate(s) but rather to the ability of these species to grow and undergo fibril formation, which depends on the presence of monomeric Aβ42. These findings contribute significantly to the understanding of amyloid formation and toxicity in Alzheimer disease, provide novel insight into mechanisms of Aβ protofibril toxicity, and important implications for designing anti-amyloid therapies.

during the past 3 decades have focused on elucidating the mechanisms of A␤ fibrillization, identifying toxic species, and developing strategies to inhibit and/or reverse A␤ amyloid formation and toxicity in vivo (3,4).
A␤ peptides are produced as soluble monomers (5,6) and undergo oligomerization and amyloid fibril formation via a nucleation-dependent polymerization process (7,8). During the course of in vitro A␤ fibril formation, various nonfibrillar aggregation intermediates, collectively called soluble oligomers or protofibrils, have been shown to precede the emergence of fibrils. Increasing evidence from various sources points to A␤ oligomers/protofibrils as putative toxic species in AD pathogenesis and suggests that these species are potential therapeutic targets for treating AD (reviewed in Refs. 9,10). Although the toxic oligomer hypothesis has emerged as one of the major current working hypotheses in AD research, the development of effective diagnostic tools and therapies on the basis of this hypothesis has yet to be realized (11)(12)(13). This is partially due to the fact that identification of a single toxic A␤ species that correlates with AD progression and severity remains elusive. Furthermore, the exact mechanisms by which these species contribute to A␤ toxicity in vivo and the nature of the toxic species are not yet fully understood. Recent evidence suggests that accelerating the process of A␤ fibrillization greatly enhances A␤ toxicity in vitro (14) and the spread of amyloid pathology in vivo (15)(16)(17).
Despite significant efforts by different groups to isolate specific intermediates along the amyloid formation pathway (12, 18 -22), the inherent heterogeneity of the process and metastable nature of A␤ oligomers (11)(12)(13) have precluded the isolation of a single toxic species. Unless covalently crosslinked (23), A␤ oligomers do not exist as stable entities, i.e. they evolve into higher order aggregates and, if they are onpathway intermediates, convert into fibrils (19). Therefore, it is plausible to assume that the structural dynamics of oligomers and factors that govern their interconversion and/or growth might influence some of the disease-related cytotoxic effects of A␤. In other words, an ongoing polymerization process involving the elongation and growth of oligomers, rather than the formation of a stable oligomeric species, may be responsible for A␤ toxicity and neurodegeneration in AD.
To test this hypothesis, we developed different procedures to isolate A␤ species of defined size and morphology distribution (24), and we investigated their toxicity in different cell lines and primary neurons. We observed that crude A␤42 preparations, containing a monomeric and heterogeneous mixture of A␤42 oligomers and protofibrils, were more toxic than the purified monomeric protofibrillar fractions or fibrils. The toxicity of protofibrils was directly linked to their interactions with monomeric A␤42 and strongly dependent on their ability to convert into amyloid fibrils.
Selective removal of the monomers, by SEC or by degradation with insulin-degrading enzyme (IDE), retarded the elongation of protofibrils, their fibrillization, and diminished protofibril toxicity toward cultured rat primary neurons, pheochromocytoma (PC12) cells, and neuroblastoma (SHSY5Y) cells. Similarly, we show that an ongoing A␤42 polymerization process, rather than distinct A␤42 aggregate states, also underlies previously reported alterations in astrocyte metabolic phenotypes (25). These findings contribute significantly to the understanding of amyloid formation and propagation in AD, provide novel insight into the mechanisms of A␤ protofibril toxicity, and carry important implications for designing anti-amyloid therapies.

EXPERIMENTAL PROCEDURES
Unless indicated otherwise, chemicals and reagents of analytical grade were purchased from Sigma. A␤ peptides were synthesized and purified by Dr. James I. Elliott, Yale University, New Haven, CT, as described previously (26). Chromatography columns (Table 1) were purchased from GE Healthcare except TSK-GEL G4000 PW XL (TSK4000), which was purchased from Tosoh Bioscience (Belgium). Highly purity distilled water was used to prepare buffers, and solutions were filtered and degassed by passing through vacuum-driven 0.22-m stericup filtration units (Millipore, Switzerland) before use. Deoxy-D-glucose (2-[1,2-3 H]glucose (2-[ 3 H]DG), specific activity, 30 -60 Ci/mmol) was obtained from ANAWA (Switzerland). DNase and papain were purchased from Sigma (catalog no. D4527 and P4762, respectively).

Preparation of A␤42 Protofibrils
Crude A␤42 protofibril (A␤42 CR) solution was prepared as described previously (24,27). Briefly, 1 mg of lyophilized A␤42 was solubilized in 50 l of 100% anhydrous DMSO in a 1.5-ml sterile microtube. Then 800 l of high purity water was immediately added, and the pH was brought to ϳ7.6 by adding 10 l of 2 M Tris base, pH 7.6. The solution was always freshly prepared and used immediately.

Size Exclusion Chromatography
Separation of A␤ Protofibrils and Monomers-SEC fractionation was carried out using an Ä KTA Explorer FPLC (GE Healthcare) placed inside a cold (4°C) chamber. SEC columns were thoroughly equilibrated with SEC running buffer (10 mM Tris-HCl, pH 7.4) prior to A␤ injections. A␤ (40 and 42) monomers and A␤42 protofibrils were obtained by SEC as described previously (24,27). To prepare A␤ monomers, 1 mg of lyophilized A␤ was dissolved in 1 ml of 6 M guanidine HCl solution and subsequently centrifuged (16,000 ϫ g, 4°C, 10 min). The supernatant was injected into a Superdex 75 HR 10/30 column, and A␤ was eluted at flow rate of 0.5 ml/min (1 ml/fraction). A␤ elution was monitored at absorbance wavelengths of A 210 , A 254 , and A 280. To obtain both protofibrils (PF) and monomers (M), A␤42 crude (CR) solution was prepared as described above. After centrifugation (16,000 ϫ g, 4°C, 10 min), the supernatant was fractionated on a Superdex 75 column as described above. The fractions eluting in the void volume were combined and labeled as protofibrils, whereas the fractions eluting under the 11-13-ml peak were combined and labeled as monomers. The PF and M fractions were further characterized by thioflavin-T (ThT) binding and transmission electron microscopy (TEM) as described below and used immediately for analytical SEC and toxicity studies.
Subfractionation of A␤42 Protofibrils-A␤ CR solution was prepared and centrifuged as described above. The supernatant was injected into either a single SEC column with a higher molecular weight separation range than Superdex 75 or a combination of SEC columns connected in series (Table  1) and fractionated as described above. For experimental purposes, the fractions corresponding to monomeric elution were combined. However, the fractions corresponding to protofibrillar fractions were kept separate and labeled according to their elution position ( Table 1). The fractions were further characterized by ThT binding and TEM and used immediately for analytical SEC and toxicity studies.
A␤ Concentration Determinations-A␤ concentration in fractions was determined by UV absorbance at 280 nm (A 280 ) using the theoretical molar extinction coefficient 1490 M Ϫ1 cm Ϫ1 (28).
Preparation of Fibrils and Sonicated Fibrils-A␤42 F were prepared by incubating the CR solution (200 M) at 37°C with gentle shaking for 24 -48 h. Fibril formation was confirmed by ThT binding and TEM. To generate fibril seeds (SF), 100 -300-nm-long fibrillar structures, the fibrils were sonicated on ice using a Vibra-Cell TM instrument (Sonics and Materials, Inc.) equipped with a fine tip (five times, 20-s pulses, amplitude 40%, output watts 6, 20-s delay between successive pulses). Subsequently, the quality of SF preparation and fibril fragmentation was confirmed by TEM.
Analytical SEC (Ana-SEC)-Ana-SEC was carried out to assess the efficiency of fractionation and heterogeneity in

Cell Culture Toxicity Studies
Primary Neuronal Cell Cultures-Primary cortical neurons were prepared from Sprague-Dawley rats (Charles River Laboratories, L'Arbresle, France) at postnatal day 1 essentially as described previously (29). After removing the meninges, the cortical tissue was cut into small pieces and enzymatically disrupted in dissociation buffer (papain, CaCl 2 , EDTA, and HEPES; 30 min; 37°C). The DNase was added for 10 min. Cortex pieces were then washed three times in prewarmed complete growth medium (neurobasal ϩ 10% FCS ϩ penicillin/streptomycin ϩ L-glutamine; Invitrogen) and then triturated seven times in 5 ml of medium. Cells were filtered through a 0.45-m cell strainer, counted, and plated onto 96-well, poly-L-lysine-precoated plates (30,000 cells/200 l/ well). After 90 min of incubation at 37°C, the medium was replaced by astrocyte-conditioned growth medium. On the 4th day in vitro (DIV 4), 2.5 M cytarabine (Sigma) was added.

A␤ Toxicity Studies in Primary Neuronal Cultures
Primary neurons were treated with the following A␤42 preparations: 1) A␤42 crude; 2) A␤42 fibrils; 3) SEC purified protofibrils and monomers; and 4) 1:1 molar mixtures of protofibrils and monomers. For this purpose, on DIV 7 half of the culture medium was replaced by complete growth medium containing A␤ preparations (0.2 (v/v) medium dilution; A␤ preparations and the buffer vehicle contained 140 mM NaCl and 10 M final A␤ concentration). After 24 h of treatment, neuronal cell viability was assessed by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) reduction assay as described below.

A␤ Toxicity Studies in PC12 and SHSY5Y Cell Cultures
Toxicity of A␤42 Crude, Fibrils, Protofibrils, and Monomers-After 24 -48 h in culture, half of the culture medium was removed and replaced with the indicated A␤ preparations. For this purpose, A␤ preparations were added to the supplemented 10ϫ DMEM (Table 2) at a 9:1 (v/v) ratio. PC12 and SHSY5Y cells were treated with final A␤ concentration of ϳ10 M. After 24 h of treatment, cell viability was assessed by MTT reduction assay. Medium aliquots were obtained for ThT binding and TEM.  (Table 1), and sonicated fibrils were prepared in supplemented 10ϫ DMEM (Table 2). Briefly, the following mixtures were prepared (molar ratios): 1) protofibrils and monomers (1:1); 2) protofibrils and sonicated fibrils (4:1); 3) monomers and sonicated fibrils (4:1); 4) protofibrils, monomers, and sonicated fibrils (2:2:1); and 5) protofibrils and monomers, and the monomers were incubated with insulin-degrading enzyme (R & D Systems catalog no. 2496-ZN) for 20 min at room temperature (A␤: recombinant human IDE, 20:1) and then added to protofibrils. PC12 and SHSY5Y cells were treated with mixtures 1-5 (final A␤ concentration of Ϸ10 M), and after 24 h, viability was assessed by MTT reduction assay. Medium aliquots were obtained for ThT binding and TEM.
Time Course of A␤42 Toxicity-PC12 cells were cultured in phenol red-free DMEM (Invitrogen catalog no. 21063-029) supplemented with 1% penicillin/streptomycin (Invitrogen) and 2 M human recombinant insulin (Invitrogen). PC12 cells were treated with 10 M A␤42 crude solution. Medium aliquots were obtained at selected time points and analyzed by ThT binding and TEM. In parallel, cell viability was assessed by MTT reduction and lactate dehydrogenase release in the media as described below.

Cell Viability Assays
MTT Reduction Assay-The viability of A␤-treated cells was assessed by MTT reduction assay using a commercial kit (Promega, catalog no. G4000) according to the manufacturer's instructions. The absorbance was measured at 570 nm ( 690 nm was used as a reference wavelength) in a Safire2 microplate reader (TECAN). The signal was expressed as percentage of the A 570 -690 count from the vehicle-treated cells (100%).
Lactate Dehydrogenase Release Assay-The lactate dehydrogenase release assay was carried out on medium aliquots using a commercial kit (Sigma TOX-7) according to the manufacturer's instructions. The lactate dehydrogenase content in the medium of the vehicle treated cells was expressed as 100%.

2-[ 3 H]DG Utilization-2-[ 3 H]
DG utilization experiments were conducted as described previously (25). The media used for 2-[ 3 H]DG uptake were pre-equilibrated at 37°C, 5% CO 2 and 95% air. To initiate the 2-[ 3 H]DG assay, the medium was replaced by 2 ml of serum-free medium (DMEM (D5030, Sigma) supplemented with 5 mM glucose, 44 mM NaHCO 3 , and 1% of antibiotic/antimycotic solution) containing 1 Ci/ml 2-[ 3 H]DG. The cells were incubated for 20 min at 37°C, 5% CO 2 and 95% air. The assay was terminated by aspiration of the culture medium and washing the cells three times with 4 ml of ice-cold PBS. Cells were then lysed by adding 2 ml of 10 mM NaOH containing 0.1% Triton X-100. Then 500-l aliquots, in duplicates, were assayed for radioactivity using liquid scintillation counting. The results were calculated by subtracting from the total counts the portion that was not inhibited by the glucose transporter inhibitor cytochalasin B (25 M), which was added 20 min prior to and along the 2-[ 3 H]DG incubation. 2-[ 3 H]DG utilization was calculated as femtomoles per dish and then expressed as percentage of the vehicle control.
ThT Binding Assay-ThT binding assay was performed by mixing 80 l of A␤ preparations (in SEC buffer or cell culture media) with 20 l of ThT (100 M) and 10 l of glycine-NaOH, pH 8.5 (500 mM), in a Nunc 384-well fluorescence plate (100 l/well). The ThT fluorescence of each sample was measured in an Analyst AD fluorometer (Molecular Devices, Switzerland) at excitation and emission wavelengths of 450 nm and 485 nm , respectively. TEM Sample Grid Preparation and Image Acquisition-A 5-10-l droplet of sample containing A␤ (A␤ preparations or cell culture media) was deposited on a 200 mesh Formvarcoated TEM grid (EM Sciences) and was allowed to settle for ϳ60 s. The excess solution was wicked away by gently applying a piece of blotting paper to the edge of the grid. Then a 10-l droplet of 2% uranyl acetate was deposited on the grid and allowed to settle for ϳ60 s. The excess solution was removed as above. Finally, the grid was vacuum-dried by gently applying the vacuum probe close to the grid edges. Image acquisition was carried out using a Phillips CM10 microscope operated at an acceleration voltage of 80 kV.

SEC Subfractionation Retards Elongation of Protofibrils and
Decreases Toxicity-We and others have shown that toxic A␤ protofibril preparations consist of a heterogeneous mixture of high molecular weight aggregates of various sizes and morphologies (12,24,30). To determine whether the cytotoxicity of protofibrils is associated with specific type of aggregates or depends on the dynamics of inter-conversion among different aggregate species, we subjected a crude A␤42 protofibril preparation (A␤42 CR) to fractionation by SEC (12,24). Then we investigated the cellular toxicity of individual A␤42 fractions, enriched with distinct protofibrillar aggregates, using cultured PC12 and SHSY5Y cells. Subfractionation of A␤42 protofibrils on a Superose 6 column resulted in 4 -6 fractions corresponding to the elution of protofibrils (Fig. 1A). Only the first four protofibril fractions (F1-F4) had sufficient A␤ content (Ն25 M) to be useful for toxicity experiments. The monomers eluted later in two fractions (F5 and F6) that were separate from protofibrils (Fig. 1A). Reinjection of the purified A␤ species onto an Ana-SEC column showed that only F1 and F4 exhibited distinct elution profile for the protofibril peak, suggesting marked differences in the size distribution of the aggregates in these fractions (Fig. 1B). Soon after SEC fractionation, F1 consisted of large aggregates and clusters of protofibrils (Fig. 1F), whereas F3 was predominantly enriched with short curvilinear and spherical protofibrils (Fig. 1G). The monomer fraction (F5) was devoid of any visible aggregates (Fig. 1H). Ana-SEC of the monomer fractions (F5 and F6) showed a single narrow elution peak (data not shown).
Cell viability assessments of A␤-treated cultured PC12 and SHSY5Y cells revealed that crude A␤42 protofibrils (CR) decreased viability by ϳ40 -45% (p Ͻ 0.01), whereas the toxicity of A␤42 SEC fractions (protofibrils and monomers) did not exceed ϳ10 -20% (Fig. 1C). A␤40 monomers were also tested as control and consistently showed only a slight, but not significant, effect on the cell viability (ϳ5-10%, p Ͼ 0.05; Fig.  1C). To determine whether the observed toxicity correlated with the changes in A␤42 aggregation state in individual fractions, we sampled the cell culture medium at different time points and probed the aggregation state of A␤ using ThT and TEM. Only the crude protofibrils exhibited a significant and gradual rise in ThT binding and reached a maximum within 24 -48 h. In contrast, none of the purified protofibril fractions (F1-F4) showed a comparable rise in ThT signal even after 96 h of incubation. ThT binding by the monomer fraction (F5), after 96 h of incubation, did not exceed that of the protofibrils (Fig. 1D).
On the basis of these observations, we hypothesized that the loss of fibrillization led to attenuation of A␤ toxicity and that the toxicity in crude A␤42 protofibrils (CR) was mediated by an ongoing process of A␤ polymerization rather than the formation of specific/stable aggregation state(s). In support of this hypothesis, analysis of the culture medium by TEM revealed that only the crude A␤42 protofibrils (CR) formed extensive mature fibrils (Ͼ1 m long and high ThT binding) during the course of the experiments (Fig. 1I). Even after 96 h of incubation, mature fibrils were not observed in culture medium containing the purified protofibril fractions F1 and F3 (Fig. 1, J and K, respectively). Instead, clusters of elongated curvilinear protofibrils were observed. The monomer fraction (F5) did show some fibril-like structures (Fig.  1L), but they were not as extensive as those seen in the case of the crude A␤42 protofibrils (Fig. 1I). Subsequent subfractionation experiments, using combinations of SEC column connected in series (Table 1), confirmed these findings and revealed that the fibrillization and toxicity of crude A␤42 protofibrils were closely linked processes (supplemental Figs. 1 and 2).
Separation of Monomers from Protofibrils Retards A␤42 Fibril Formation and Reduces the Toxicity of Protofibrils-Some of the possible explanations for the reduced fibrillization and toxicity of A␤42 protofibrils upon SEC subfractionation include the following: 1) the presence of a small number of fibrils in the crude preparations that are lost upon SEC fractionation; 2) disruption of the structure of toxic A␤ species upon interaction with the SEC column matrices; 3) disruption of the dynamics of inter-conversion among different protofibril species; or 4) retardation of protofibril growth due to the removal of excess free A␤ monomers. To investigate these possibilities, we compared the aggregation and toxic properties of crude A␤ protofibrils (CR) with those of monomers (M), fibrils (F), and the protofibrils (PF) that were not subfractionated and simply separated from monomers. When A␤42 CR was fractionated on a Superdex 75 SEC column, all protofibrils eluted in the void volume without further separation (cutoff Ն70 kDa) and monomers eluted later in the included volume ( Fig. 2A), as reported previously (27,30). We have verified that these conditions yield protofibrils and monomer fractions that are free of fibrils (24). The PF fraction contains a heterogeneous mixture of different quaternary structures. Cell viability assessments of A␤-treated primary neuron cultures, PC12 cells, and SHSY5Y cells revealed that 24 h of treatment with A␤42 CR decreased the viability of PC12 and SHSY5Y cells by Ն40% (p Ͻ 0.01) and of primary neurons by ϳ50% (p Ͻ 0.01) (Fig. 2C). The purified, but not subfractionated, PF decreased cell viability by ϳ20 -30% (p Ͻ 0.05), and A␤42 M exhibited only a slight effect (ϳ5-20%, p Ͼ 0.05) (Fig. 2C). Treatment of cells with preformed F, obtained by incubation of monomers at the same concentration, also impaired cell viability by ϳ10 -20% (p Ͼ 0.05) but was remarkably less toxic than crude protofibrils or purified PF.
As observed earlier (Fig. 1D), the A␤42 CR protofibrils exhibited substantial rise in ThT binding within 24 h (Fig. 2D). As a control, mature preformed F were also added to the culture medium and showed a ThT signal similar to crude A␤42 protofibrils at 24 h. These observations confirmed the substantial fibrillization of A␤42 CR during the time interval of toxicity experiments (24 h). TEM analysis of culture medium revealed that A␤42 CR preparations formed extensive mature fibrils within 24 h (Fig. 2I, compare with the image of preformed fibrils in Fig. 2H). Neither M nor purified PF exhibited a comparable rise in ThT binding over time (Fig. 2D). This is consistent with TEM observations that revealed the absence of mature fibrils in the culture medium and only clusters of protofibrils after 24 h of incubation (Fig. 2, J and K,  respectively).
Addition of Monomers to Protofibrils Enhances A␤42 Fibrillization and Toxicity-Our results suggest that SEC fractionation of protofibrils reduces, but does not completely eliminate, their toxicity and demonstrate that toxicity of A␤42 protofibrils is strongly linked to their ability to undergo fibrillization and form mature fibrils. Therefore, we wanted to investigate if the slow fibrillization of protofibrils, and thus their reduced toxicity, were due to the removal of excess monomers and/or fibrils by SEC. To this end, we reintroduced A␤42 monomers, or preformed fibrils, to purified PF (obtained using a Superdex 75 column) or subfractionated protofibril fractions (F1-F4, obtained using a Superose 6 column), and we assessed the fibrillization and toxicity of each fraction and mixtures thereof. We hypothesized that the addition of free monomers and/or preformed fibril seeds would enhance both protofibril fibrillization and toxicity. Accordingly, we observed that reintroduction of A␤42 monomers to purified PF resulted in an increase in fibril formation and enhanced protofibril toxicity (supplemental Fig. 2). To confirm these findings, we added purified monomers to the subfractionated A␤42 protofibrils (F1-F4). Mixtures of monomers and protofibril fractions (F1-F4) exhibited higher ThT binding over time (Fig. 3A) as compared with fractions containing protofibrils only (Fig. 1D). However, it became apparent that the extent of ThT binding in the monomer/protofibril mixtures was not as robust as seen in the case of A␤42 CR (Fig. 1D), suggesting a weak seeding capacity of subfractionated protofibrils. Treatment of cells with mixtures of A␤42 monomers and subfractionated protofibrils (F1-F4) impaired cell viability by ϳ25-35% (Fig. 3B, p Ͻ 0.05), which was comparable with that seen with A␤42 crude (Ն40%, Fig. 2C). TEM analysis of the culture medium revealed that addition of A␤42 monomers to protofibrillar fractions (F1-F4) induced substantial fibril formation (supplemental Fig. 3, A and B, e.g. F1ϩM and F3ϩM, respectively, data for F2ϩM and F4ϩM is not shown). The monomer fraction (F5), after 96 h of incubation, showed the presence of weakly ThT binding immature fibrils ( Fig. 3A and supplemental Fig. 3C).
Taken together, if A␤42 fibrillization and toxicity are closely linked processes, and if these processes require the availability of A␤ monomers, then selective degradation of monomers should significantly retard protofibril elongation and reduce their toxicity. To examine this hypothesis, we sought to induce selective removal of A␤42 monomers using IDE, an enzyme that selectively degrades A␤ monomers but not oligomers or fibrils (31)(32)(33). IDE was added to a solution of A␤42 monomers, and this mixture was then added to subfractionated protofibril fractions (F1-F4) in supplemented 10ϫ DMEM (see under "Experimental Procedures"). As anticipated, the addition of IDE reversed the effect of A␤42 monomers in enhancing protofibril toxicity (compare Fig. 3,  C with B). The toxicity of protofibrils (F1-F4) mixed with IDE-treated A␤42 monomers was similar to the toxicity seen with protofibrillar fractions (F1-F4) alone (compare Fig. 3C with 1C). Similarly, we observed that addition of IDE to toxic crude A␤42 preparations resulted in selective degradation of the monomers and diminished the cytotoxicity and fibrillization of these preparations, albeit to a lesser extent than monomer/protofibril mixtures (supplemental Fig. 4).
Next we wanted to determine whether the presence of a small amount of fibrillar aggregates in the crude preparation, which might have been lost upon SEC fractionation, also contributed to the enhanced toxicity of this preparation relative to the SEC-purified protofibrils and monomers. To investigate this, we introduced a small amount (2 M) of sonicated A␤42 fibrils (SF) to subfractionated A␤42 protofibrils (F1-F4). We found that the addition of sonicated fibrils did not lead to any significant enhancement of protofibril toxicity (Fig. 3D), regardless of differences in the size distribution between the various protofibril fractions. TEM analysis revealed that, even after 96 h of incubation, these fractions predominantly contained mixtures of fibrillar seeds and protofibrils (supplemental Fig. 3, D (F1ϩSF) and E (F3ϩSF)). In contrast, addition of SF to the monomer fraction (F5) resulted in significantly increased amyloid formation as evidenced by the presence of mature fibrils (supplemental Fig. 3F), but the toxicity was relatively unaffected (Fig. 3D). These findings may indicate that monomer-protofibril interactions, rather than monomer-fibril interactions, are the key determinant of A␤42 toxicity. Interestingly, the addition of the mixtures of M and fibrillar seeds (SF) to the protofibril fractions, and not fibrillar seeds (SF) only, induced robust toxicity in subfractionated protofibrils (supplemental Fig. 5B). To determine whether the observed findings are specific to A␤42, we carried out fibrillization and toxicity experiments using different preparations of A␤40 (crude, monomers, protofibrils, and mixtures thereof). The data show that at the concentrations similar to those used in the A␤42 (5-10 M) studies, we did not observe significant difference in terms of toxicity between monomeric, protofibrillar, or crude preparations of A␤40 (supplemental Fig. 6). These findings are consistent with the lack of changes in the ThT signal and fibrillization of these samples, thus supporting our hypothesis that A␤ toxicity is also linked to its ability to undergo fibril formation. Consistent with this hypothesis, significant toxicity (30 -40%) was only observed when A␤40 undergoes fibrillization (Ն50 M). These findings point to the processes of protofibril-monomer interactions during maturation and elongation of A␤ protofibrils/oligomers, rather than relatively stable and discrete oligomers species, as the primary processes that mediate A␤ toxicity.
Viability of the Cultured Cells Is Critically Impaired before the Emergence of Mature A␤42 Fibrils-To better understand the relationship between the process of amyloid formation and A␤42 toxicity, we treated cultured PC12 cells with A␤42 CR and sought to correlate the A␤42 aggregation state in the culture medium with cell viability at different time points. Fig. 4A shows that A␤42 CR exhibited a gradual rise in ThT binding within the first 9 h and then remained relatively unchanged. Interestingly, a gradual and early decline in MTT reduction by A␤42 CRtreated cells was also observed during the initial 6 h (Fig.  4B, solid line, Ն30% decrease compared with time 0, p Ͻ 0.01), whereas a significant rise of lactate dehydrogenase in the culture medium (Ն80%, p Ͻ 0.005 compared with time 0) was only observed after ϳ6 h of incubation (Fig. 4B,  dashed line). TEM analysis of the culture medium at ϳ3 h showed clusters of protofibrils and spherical aggregates (Fig. 4C). At ϳ6 h, the protofibrils were more elongated and had an immature fibril-like appearance (Fig. 4D). Mature fibrils started to appear at ϳ9 h (Fig. 4E) and were more abundant at ϳ24 h (Fig. 4F). These observations suggested that impairment of cell viability, indicated by a decline in the reduction of MTT, a marker for mitochondrial metabolism (34), was affected by the process of A␤42 protofibril elongation, in the presence of free monomers, rather than the formation of mature fibrils. In other words, toxicity is mediated by nonfibrillar A␤42 aggregates formed as a result of an ongoing process of protofibril elongation, whereas the emergence of mature, high ThT binding fibrils is a later event.
Crude A␤42 Preparations, but Not Monomers, Protofibrils, or Fibrils Alone, Increased Glucose Utilization by Cultured Astrocytes-We have previously reported that treatment of cultured astrocytes with A␤42, at comparable concentrations as used for cell viability measurements in the experiments described above, does not induce significant cell death but alters the astrocyte metabolic activity (25). We observed that A␤42 significantly enhances glucose utilization (uptake and phosphorylation), as assessed by 2-[ 3 H]DG uptake, by cultured astrocytes as compared with A␤40 (25). We sought to investigate whether this phenomenon was attributable to particular A␤42 assembly state(s) i.e. monomers, protofibrils, or fibrils, and/or was influenced by the dynamics of the A␤42 aggregation process. Treatment of cultured astrocytes with defined A␤ aggregates (fibrils and SEC-purified monomers and protofibrils) had a negligible effect on glucose utilization by astrocytes (Fig. 5). However, treatment with an A␤42 CR preparation, which contained substantial amounts of protofibril-abundant monomers, caused a significant (p Ͻ 0.005) increase in glucose utilization (Fig. 5). These observations suggest that the dynamics of the A␤42 fibrillization process bear important implications for not only neuronal viability but also neuron-glia metabolic coupling in AD pathogenesis.

DISCUSSION
The identification of toxic A␤ species and/or the process of their formation is crucial for understanding the mechanism(s) of A␤ neurotoxicity in AD and development of effective diagnostic and therapeutic interventions (35). Several lines of evidence implicate prefibrillar A␤ oligomers, including protofibrils, as the primary neurotoxins in AD (9,36). Aiming to identify specific toxic A␤ species, we and others have developed reproducible protocols to isolate prefibrillar aggregates of defined size and morphology distributions (22,24,30). Herein, we extended our separation protocols to further subfractionate the prefibrillar oligomers and evaluated their fibrillization and toxicity compared with the monomeric and fibrillar preparations of A␤42. We demonstrate that the crude A␤42 preparations, containing monomeric and heterogeneous mixture of A␤42 oligomers, are more toxic than purified monomeric, protofibrillar fractions or fibrils. Subfractionation of protofibril preparations, to separate different protofibril species, resulted in further attenuation of their toxicity (Figs. 1 and 2). The diminished toxicity of purified and subfractionated A␤42 protofibrils was strongly linked to their reduced fibrillization, thus indicating that A␤42 toxicity is not necessarily linked to specific prefibrillar aggregate(s) but rather to the ability of these species to grow in the presence of monomeric A␤42. Consistent with this hypothesis, reintroduction of monomeric A␤42 into these fractions restored amyloid formation and enhanced their toxicity to comparable levels as observed with the crude preparations. Furthermore, selective removal of monomeric A␤42 from these preparations using IDE reversed the toxicity of A␤42 protofibrils (Fig. 3). In parallel, we demonstrate that the crude A␤42 preparations were more potent in inducing metabolic alterations in cultured astrocytes as compared with the purified monomers, protofibrils, and fibrils (Fig. 5). Finally, we also demonstrate that promoting the fibrillization of A␤40, the major A␤ variant in vivo, also results in increased cytotoxicity to the cultured cells (supplemental Fig. 6). Together, these findings provide strong evidence that A␤ toxicity is linked to an ongoing A␤ polymerization process and is greatly reduced when the polymerization process is slowed down by the selective removal/degradation of A␤ monomers (Figs. 1-3).
Monomer-Protofibril Interactions Are Key Determinants of A␤42 Fibrillization and Toxicity-Consistent with a nucleated polymerization process, the fibrillization of A␤42 protofibrils, or lack thereof, was strongly influenced by the availability of monomers. Cell toxicity experiments revealed that A␤42 toxicity was enhanced or reduced depending on the fibrillization propensity of the protofibril or monomer preparations (Figs. 1-3). This is in agreement with observations that A␤ variants that exhibit a high aggregation propensity, e.g. A␤42 or A␤-arctic, also induce greater toxicity in cultured neurons as compared with slower aggregating A␤ variants, e.g. A␤40 (37)(38)(39). Similarly, accelerating the fibrillogenesis of soluble A␤40 preparations by adding exogenous A␤40 fibrils also enhances their toxicity (14). Our data suggest that the fibrillization and toxicity of A␤42 protofibrils is dependent on the presence of monomers and not on fibrillar forms of A␤. The presence of fibrillar seeds contributes, but is not essential, to the toxicity seen with mixtures of A␤42 protofibrils and monomers (supplemental Fig. 5).
Conversion of A␤ protofibrils into fibrils occurs by incorporation of monomers into the growing ends. Therefore, removal of monomers from solution is expected to enhance protofibril stability and reduce the rate of their fibrillization. Our findings suggest that kinetic stabilization of protofibrils would reduce their toxicity. This hypothesis is also supported by the observations that enhancing the kinetic stability of A␤42 protofibrils by the addition of A␤40 monomers significantly reduces their toxicity toward cultured neurons (27). Similarly, small molecules that stabilize A␤ protofibrils in vitro have also been shown to improve behavioral performance in APP transgenic mice (40). The critical requirement for A␤ monomers to maintain A␤ fibrillization, and associated toxicity in vitro and in vivo, is also underscored by the overexpression of IDE in APP transgenic animals. These studies revealed that overexpressing IDE in these animals ameliorates plaque pathology, improves behavioral performance, and prevents premature death (32,41). Therefore, the concentration of monomeric A␤ plays important roles in modulating the amyloid formation and toxicity of A␤ peptides. Intriguingly, we found that IDE was less efficient at improving the survival of A␤42 crude treated cells as compared with the cells treated with monomer/protofibril mixtures (compare Fig. 3C with supplemental Fig. 4B). We speculate that this can possibly result from the preferential incorporation of A␤42 monomers into protofibrils and/or binding of the IDE to A␤ oligomers but less efficient degradation due to the secondary structure elements (42).
The exact mechanisms by which the process of amyloid formation contributes to A␤ toxicity and neurodegeneration in AD remain unknown. Our results suggest that the formation of the toxic A␤ species could occur on site and is mediated by other cellular factors that interact with, or mediate the interactions between, protofibrils and monomers or respond to their fibrillization. Given that A␤ species were administered extracellularly, it is plausible to postulate that fibrillization events take place on the cell membrane where the process of amyloid formation triggers downstream intracellular cascade of events culminating in cell death. This hypothesis is supported by the findings that association of A␤ to membrane gangliosides also promotes A␤ fibrillization on the cell surface and induces toxicity (43). Alternatively, it is also possible that certain cell surface receptors are activated due to the binding of protofibrils (44) or their fibrillization, which triggers cellular pathways compromising neuronal survival capabilities.
In addition to A␤40, Wogulis et al. (14) also demonstrated that mixtures of soluble (nonfibrillar) and insoluble (fibrillar) fractions of IAPP were more toxic to the cultured neurons than fibrillar and nonfibrillar fractions alone. A recent compelling model for IAPP toxicity directly links the processes of amyloid aggregation on membrane surfaces with cell-membrane disruption and the onset of toxicity (45,46). In other words, as the fibril develops on the membrane surface, the structural integrity of the membrane is simultaneously compromised. In an elegant study, Engel et al. (45) demonstrated the synchronization of the kinetic profiles for human IAPPfibril growth, which was monitored by ThT fluorescence, and the induction of dye leakage from coincubated mixed 1,2dioleoyl-sn-glycero-3-phosphocholine/1,2-dioleoyl-snglycero-3-phospho-L-serine vesicles. The two profiles were characterized by a lag phase of ϳ3 h followed by a sigmoidal transition to human IAPP fibrils and near-complete dye leakage from the vesicles. This result indicates that it is the process of fibril formation on membrane surfaces that is responsible for abolishing the membrane barrier function. Cryogenic electron microscopy images of 1,2-dioleoyl-sn-glycero-3phosphocholine/1,2-dioleoyl-sn-glycero-3-phospho-L-serine large unilamellar vesicles coincubated with human IAPP showed distortion and pinching of regions of the membrane in contact with fibrils, whereas vesicles incubated in the presence of nonamyloidogenic mouse IAPP remained unperturbed (45). As a mechanistic model, the authors proposed that IAPP-fibril growth on the membrane occurred concomitantly with a forced change in membrane curvature and weakened lipid packing, which enabled the leakage of intravesicular contents. Interestingly, this notion is consistent with our findings, which indicate that it is not one specific oligomeric state that induces cell death and toxicity but rather the dynamic process of fibril formation.
Alternatively, the uptake and internalization of soluble A␤ (monomers and protofibrils) may impair cellular energy metabolism (intracellular toxicity). This is supported by some recent reports demonstrating that A␤ in culture medium is internalized and aggregated intracellularly into fibrils, eventually causing membrane disruption (47,48). Further studies are required to dissect the structural and molecular mechanisms underlying the critical role of protofibril-monomer interactions in A␤ toxicity. In addition to enhancing neuronal vulnerability, the process of A␤ amyloid formation also seems to interfere with the regulation of glial energy metabolism as indicated by enhanced glucose utilization by astrocytes (Fig.  5). Importantly, such metabolic changes in astrocytes are associated with deleterious consequences for neuronal viability (25) and bear important implications for neuron-glia metabolic coupling (49). Thus, our results demonstrate that accelerated A␤ amyloidogenesis triggers a host of mechanisms, involving both neurons and glia that converge on triggering neuronal dysfunction and eventual demise.
Implications for Therapies in AD-The possibility that an ongoing A␤ polymerization process, rather than specific aggregate of defined size or structure, strongly determines A␤ neurotoxicity has important implications for understanding the role of A␤ aggregation in AD pathogenesis and the design of anti-A␤ therapeutics. Our results provide direct evidence linking A␤ toxicity to the growth and fibrillization of A␤ oligomers in the presence of A␤ monomers. These studies suggest that anti-A␤ therapeutic strategies, many of which are currently being tested in animal models or in clinical trials in AD patients, hold potential as promising disease-modifying interventions. These approaches are based on the following: 1) reducing the production of A␤ monomers (50 -52); 2) sequestering A␤ monomers and oligomers to promote their clearance (53,54); 3) degrading of A␤ monomers and oligomers (41,55); and 4) preventing A␤ oligomers from stably binding to the neuronal membranes (40,56).
Reducing the monomer concentration in vivo is likely to attenuate A␤ toxicity not only by preventing the formation of additional toxic oligomers but also by reducing the growth of circulating oligomers, including protofibrils, and promoting their clearance from brain. Proof-of-principle experiments clearly show that even minimal (ϳ15%) reduction in monomer A␤ production have profound effects on amyloid pathology and related deficits in APP transgenic animals (50 -52). In the absence of free monomers, the unstable nuclei would cease to grow and possibly disintegrate. Experimental evidence for the latter stipulation can be found in reports showing that sequestration of A␤ monomers using a small engineered protein (ZA␤3) (57) led to gradual dissolution of preexisting oligomers in vitro and promotes amyloid clearance in vivo (57). The administration of therapeutic anti-A␤ antibodies that sequester A␤ monomers and oligomers or A␤ vaccination to generate such antibodies in the host prevents amyloid formation (53,58), reduces plaque burden in older plaque-bearing mice (54,58), and improves behavioral performance (59,60).
Development of small molecules that inhibit amyloid formation by stabilizing the monomeric state of A␤ has not been successful due to the conformational heterogeneity and flexibility of the protein. Our findings suggest that small molecules targeting the growth of oligomers/protofibrils and fibrils, by preventing monomer incorporation into the growing ends of these species, would constitute a more effective strategy to block and/or reverse amyloid formation and toxicity in vivo. Blocking oligomer growth by capping the ends or binding to their surfaces can be achieved at substoichiometric levels (27), whereas molecules that stabilize the monomers to prevent self-assembly would have to be used at stoichiometric amounts.
Some studies have shown that A␤ immunotherapy improves behavioral performance in APP transgenic animals without decreasing total (soluble and insoluble) A␤ burden (59,61). Although the latter observations seem to challenge the relevance of the A␤ amyloid cascade hypothesis to AD pathogenesis, our findings serve to reconcile such discrepan-cies and suggest that the amyloid cascade hypothesis has yet to be disproved. Our data support the notion that a modest reduction in soluble A␤ (monomers and oligomers) and the disruption of further nucleation and/or polymerization events would neutralize protofibril toxicity, facilitate amyloid clearance from brain, and abrogate the spreading of amyloid pathology. In conclusion, targeting the nucleated polymerization of amyloid-forming proteins offers an exciting framework for the development of anti-amyloid therapies for a host of neurodegenerative diseases characterized by the pathological accumulation of protein aggregates, including AD, Parkinson disease, and prion diseases. We hope that these findings will stimulate further research into understanding the role of the process of amyloid formation in neurodegenerative diseases and the development of in vitro and in vivo mechanistic models to design and evaluate intervention strategies.