Oligomeric and Fibrillar Species of Amyloid-β Peptides Differentially Affect Neuronal Viability*

Genetic evidence predicts a causative role for amyloid-β (Aβ) in Alzheimer's disease. Recent debate has focused on whether fibrils (amyloid) or soluble oligomers of Aβ are the active species that contribute to neurodegeneration and dementia. We developed two aggregation protocols for the consistent production of stable oligomeric or fibrillar preparations of Aβ-(1–42). Here we report that oligomers inhibit neuronal viability 10-fold more than fibrils and ∼40-fold more than unaggregated peptide, with oligomeric Aβ-(1–42)-induced inhibition significant at 10 nm. Under Aβ-(1–42) oligomer- and fibril-forming conditions, Aβ-(1–40) remains predominantly as unassembled monomer and had significantly less effect on neuronal viability than preparations of Aβ-(1–42). We applied the aggregation protocols developed for wild type Aβ-(1–42) to Aβ-(1–42) with the Dutch (E22Q) or Arctic (E22G) mutations. Oligomeric preparations of the mutations exhibited extensive protofibril and fibril formation, respectively, but were not consistently different from wild type Aβ-(1–42) in terms of inhibition of neuronal viability. However, fibrillar preparations of the mutants appeared larger and induced significantly more inhibition of neuronal viability than wild type Aβ-(1–42) fibril preparations. These data demonstrate that protocols developed to produce oligomeric and fibrillar Aβ-(1–42) are useful in distinguishing the structural and functional differences between Aβ-(1–42) and Aβ-(1–40) and genetic mutations of Aβ-(1–42).

cascade" hypothesis has defined the fibrillization of A␤ into amyloid deposits, a pathologic hallmark of AD, as a toxic gain of function (2). That A␤-(1-42) is more fibrillogenic than A␤-  fits well with this hypothesis. However, amyloid plaques do not always correlate in number, tempo, or distribution with neurodegeneration or clinical dementia. Thus, recent debate within the AD community has focused on whether fibrillar (amyloid) or soluble oligomers of A␤ are the active species of the peptide that ultimately cause the synaptic loss and dementia associated with AD (3)(4)(5)(6)(7). In vivo, small, stable oligomers of A␤-(1-42) have been isolated from brain, plasma, and cerebrospinal fluid (8 -10) and correlate with the severity of neurodegeneration in AD (11,12). Thus, although genetic evidence predicts that A␤ is a causative factor in AD, the role of fibrillar and oligomeric A␤ in the pathogenesis of AD remains unclear.
Initial in vitro studies suggested that A␤-induced neurotoxicity required the peptide to adopt a fibrillar aggregation state, with unaggregated peptide at low doses actually exhibiting neurotrophic effects (13)(14)(15)(16)(17)(18). Recent studies demonstrate that non-fibrillar structures, including oligomers and amyloid-derived diffusible ligands (ADDLs) (19 -23), and protofibrils (24 -26) are also neurotoxic. For the present in vitro study, we describe the development of two aggregation protocols for the consistent preparation of fibrillar and small oligomeric species of A␤- . We use these preparations to directly compare the effect of oligomeric, fibrillar, and unaggregated peptide species on neuronal viability. In addition, we apply these A␤-  oligomer-and fibril-forming conditions to A␤-  to compare the effect on structure and neuronal viability with A␤- .
Several genetic mutations within the A␤ peptide sequence have been reported, including the Flemish (A21G) (27), Italian (E22K) (28), Dutch (E22Q) (29,30), Arctic (E22G) (31), and Iowa (D23N) (32,33) mutations. Clinically, these mutations lead to AD-like symptoms secondary to intracerebral hemorrhage, although in some cases a progressive dementia is present in the absence of stroke. Familial AD mutations within the APP gene induce an increase in either the total amount of A␤-(1-42) (Swedish) (34) or the ratio of A␤(1-42)/A␤(1-40) (London) (35) by affecting the activity of ␤and ␥-secretase. In contrast, the 19 -25 region of the A␤ peptide is critical to several properties of the peptide (36), including proteolysis by ␣-secretase (33,37), aggregation (38,39), and neurotoxicity (40). The Dutch and Arctic mutations are of particular interest for investigating the effect of the A␤-(1-42) oligomer-and fibril-forming conditions on structure and neuronal viability, because these mutations appear to change the aggregation of the peptide in the absence of an increase in either the total amount of A␤ or the ratio of A␤(1-42)/A␤(1-40) and actually reduce the amount of A␤-(1-42) in transfected cells ( 41,42). Specifically, the Dutch mutation is associated with an increased rate of fibrillization (43,44), and the Arctic mutation induces a significant increase in protofibril formation with no difference in fibrillization (31). In terms of toxicity, A␤-  with the Dutch mutation induced greater toxicity in vitro than wild type peptide (38,41,45). The effect of the Arctic mutation on A␤-induced neurotoxicity has not been reported as yet. Thus, the second goal of the current study was to utilize our A␤-(1-42) oligomer-and fibril-forming aggregation protocols to compare the structure and effect on neuronal viability of A␤-  with the Dutch and Arctic mutations to wild type peptide.

EXPERIMENTAL PROCEDURES
Preparation of Peptide-Synthetic and recombinant wild type A␤-(1-42) and A␤-(1-40) were purchased from American Peptide (Sunnyvale, CA) and Recombinant Peptide (Athens, GA). A␤-(1-42) E22Q was purchased from California Peptide (Napa, CA), and A␤-(1-42) E22G was purchased from Recombinant Peptide. In the standard quality control analysis we perform with every new lot of peptide, including high performance liquid chromatography and mass spectrometry, all peptides used were Ͼ95% pure.
As summarized in Fig. 1, the A␤-(1-42) peptide was initially dissolved to 1 mM in hexafluoroisopropanol (Sigma) and separated into aliquots in sterile microcentrifuge tubes. Hexafluoroisopropanol was removed under vacuum in a Speed Vac, and the peptide film was stored dessicated at Ϫ20°C. For the aggregation protocols, the peptide was first resuspended in dry dimethyl sulfoxide (Me 2 SO, Sigma) to a concentration of 5 mM. Based on one of the amyloid-derived diffusible ligand protocols developed by Krafft, Finch, Klein, and coworkers (21), for oligomeric conditions, Ham's F-12 (phenol red-free, BioSource, Camarillo, CA) was added to bring the peptide to a final concentration of 100 M and incubated at 4°C for 24 h. For fibrillar conditions, 10 mM HCl was added to bring the peptide to a final concentration of 100 M and incubated for 24 h at 37°C. For unaggregated conditions, the 5 mM A␤ in Me 2 SO was diluted directly into cell culture media. After these solubilization and aggregation protocols, no major differences were observed in the preparation and structural characterization of synthetic and recombinant peptide.
AFM-Peptide preparations were characterized using a NanoScope IIIa Scanning Probe Workstation equipped with a MultiMode head using an E-series piezoceramic scanner (Digital Instruments, Santa Barbara, CA). AFM probes were single-crystal silicon micro cantilevers with 300-kHz resonant frequencies and 42 N/m spring constant model OMCL-AD160TS-W2 (Olympus, Japan). Samples were imaged under a blanket of dry helium. Samples were prepared by spotting 10 -50 l of solution on freshly cleaved mica, which were then incubated at room temperature for 5 min, rinsed with 0.02-m-filtered deionized NANOpure water (Barnstad Thermoline, Dubuque, IA), and blown dry with tetrafluoroethane (CleanTex MicroDuster III). Image data was acquired at scan rates between 1 and 2 Hz with drive amplitude and contact force kept to a minimum.
Gel Electrophoresis-Samples were diluted in NuPage sample buffer and separated by SDS-PAGE on a 12% NuPage Bis-Tris gel (Invitrogen). The protein was transferred to polyvinylidene difluoride membranes (Invitrogen) and then blocked for 1 h in a solution of 5% nonfat dry milk in TBS/Tween 20. The membrane was then incubated with 6E10 (1:3000), a mouse monoclonal A␤ antibody to residues 1-17 (Signet, Dedham, MA). For detection, the membrane was incubated with horseradish peroxidase-conjugated Ig anti-mouse antibody (1:5000), developed using enhanced chemiluminescence (ECL, Amersham Biosciences), and visualized using an Eastman Kodak Co. 440 CF Image Station (PerkinElmer Life Sciences). Molecular mass was estimated by Rainbow molecular weight markers (Amersham Biosciences).
Cell Viability Assay-Neuro-2A cells were plated at a concentration of 5 ϫ 10 3 cells/well in 96-well plates in 100 l of media. After overnight incubation, the cells were rinsed with serum-free media containing N2 supplements (Invitrogen). Vehicle or peptide preparations were dissolved in this media and added to the cells. Staurosporine (0.5 M) was used as a positive control for apoptosis (data not shown). The treated cells were incubated for 20 h at 37°C in 5% CO 2 . As taken from the Roche Molecular Biochemicals protocol, the 3-[4,5-dimethylthizaol-2yl]-2,5-diphenyl tetrazolium bromide (MTT) reagent is reconstituted in phosphate-buffered saline to 5 mg/ml. The solubilization solution is 10% SDS in 0.01 M HCl. Both were separated into aliquots, stored atϪ20°C, and thawed upon use. 10 l of MTT labeling reagent was added to each well, and the plate was incubated at 37°C for 4 h. 100 l of solubilization solution was added to each well, and the plate was incubated overnight at 37°C. The absorbance of the samples was measured at 563 nm (Microplate Photometer, Packard Instrument Co.). For statistical analysis, an unpaired Student's t test with unequal variance was used.

RESULTS
The intent of this manuscript was to develop three distinct, reproducible structural species of wild type A␤-(1-42) (unaggregated monomers, oligomers, and fibrils) and determine their effect on neuronal viability. We then applied the conditions for producing these structural species of wild type A␤-(1-42) to A␤-(1-40) and A␤-(1-42) with either the E22G or E22Q mutations, characterized the resulting structural species, and determined the effect of these species on neuronal viability.
Oligomeric and fibrillar species of A␤-(1-42) were prepared as described above and summarized in Fig. 1. By AFM, oligomeric preparations of A␤-(1-42) appear as small globular structures that measure ϳ2-5 nm in z-height ( Fig. 2A, panel A). It is not possible to determine the number of monomers composing oligomers because their size by AFM is not homogenous. Although the precise oligomer stoichiometry remains unclear, the preparation is predominantly free of short protofibrils and entirely free of fibrils or large globular aggregates. Because the measured z-height value by AFM tends to underestimate the actual molecular diameter due to physical compression exerted by the AFM probe, measured protein z-height values of 2-5 nm would correspond to molecular masses ranging from 10 to 100 kDa (9). Fibrillar preparations of A␤-(1-42) appear as long threads measuring ϳ4 nm in z-height and Ͼ1 m in length, with numerous oligomeric species but few large globular structures ( Fig. 2A, panel B). With the A␤-(1-42) preparation protocol described, no major differences were observed in the structure of wild type synthetic and recombinant peptide; for example Fig. 2A (panels A and B) shows oligomeric and fibrillar synthetic A␤-(1-42), whereas Fig. 4A (panels A  and B) shows oligomeric and fibrillar recombinant A␤-(1-42). Applying the A␤-(1-42) oligomer-forming conditions to A␤- , the majority of the peptide remained predominantly as unassembled monomer ( Fig. 2A, panel C, inset). However, a sparse population of A␤-(1-40) oligomers measuring ϳ2-4 nm in z-height were observed, comparable with those formed by A␤-(1-42) ( Fig. 2A, panel C). A␤-(1-40) did not form extended fibrils under the A␤-(1-42) fibril-forming conditions described for these experiments, although short fibrils ( Fig. 2A, panel D) were observed in fields of otherwise predominantly unassembled peptide ( Fig. 2A, panel D, inset). Thus, the conditions we established for the formation of A␤-(1-42) oligomers and fibrils did not result in similar structures when applied to A␤- .
In addition to microscopy, immunoblotting has been used to distinguish aggregation species of A␤ (22, 24, 46). Although it is not possible to equate structural species observed by AFM with specific bands on a Western blot, differences between the oligomeric and fibrillar preparations can be observed. With SDS-PAGE, oligomeric preparations of A␤-(1-42) resolve to primarily monomer, trimer, and tetramer, whereas fibrillar preparations contain primarily monomer and large aggregates of ϳ50 -100 kDa (Fig. 2B, lanes 1 and 2). The Western blot for A␤-(1-40) treated with either the A␤-(1-42) oligomer-and fibril-forming conditions resolve to immunoreactive species consistent with A␤ monomer and tetramer, with an additional ϳ50 -100 kDa species in the A␤-(1-40) treated with the A␤-(1-42) fibril-forming conditions (Fig. 2B, lanes 3 and 4). This pattern is distinct from that of A␤-  in the lack of a prominent trimer band. Several other, minor bands are visible, but this gel analysis is included simply as a comparison of the oligomeric and fibrillar species and is not intended as a definitive identification of all the various aggregation species pres- Unlike mutations that increase either the total amount of A␤ or the ratio of A␤(1-42)/A␤(1-40), including the Flemish (A21G) mutation, the Dutch (E22Q) and Arctic (E22G) mutations actually reduce A␤-(1-42) levels in transfected cells and increase the formation of fibrils or protofibrils, respectively (31,38,41,42). Therefore, we utilized our A␤-(1-42) oligomer-and fibril-forming protocols to compare the structure and neurotoxicity of A␤-(1-42) with E22Q or E22G mutations to wild type peptide. AFM imaging revealed that under the oligomer-form-ing conditions, A␤-(1-42) with the E22Q mutation formed abundant protofibrils along with oligomers, and the E22G mutation formed fibrils measuring ϳ2 nm in z-height in addition to oligomers (Figs. 4A, panels C and E). Under fibril-forming conditions, E22Q formed fibrils that vary in diameter, with smaller sections measuring ϳ2 nm interspersed with larger ϳ5-nm sections that appeared more rigid (Fig. 4A, panel D). E22G formed fibrils that have a uniform z-height of ϳ5 nm, slightly larger and more rigid than wild type fibrils (Fig. 4A, panel F). By Western blotting, oligomer preparations of A␤-(1-42) wild type and E22Q were similar with monomer, trimer, and tetramer species, although the wild type also had aggregates of ϳ30 -40-kDa species in greater abundance than E22Q (Fig. 4B, lanes 1 and 2). The fibrillar species of wild type A␤-(1-42) contained a monomeric band and large aggregates of ϳ50 -100 kDa, whereas E22Q also had a faint monomeric band, but the majority of the fibrillar species remained in the well, too large to enter the gel (Fig. 4, lanes 3 and 4). Wild type oligomeric A␤-(1-42) and E22G migrated similarly in SDS-PAGE, comparable with E22Q (Fig. 4B, lanes 5 and 6). Fibrillar E22G contained a trimer band more abundant than in the wild type fibril preparation (Fig. 4B, lanes 7 and 8).
In terms of cell viability, oligomeric preparations of the A␤-(1-42) mutants were not consistently different from wild type oligomers (Figs. 5, A and C), although E22Q was less toxic than wild type at 100 nM, and E22G was more toxic at 100 nM but less toxic at 15 M. This is surprising because these "oligomer" preparations contained abundant amounts of protofibril (E22Q) and fibrils (E22G). This suggests that with these mutations, the effects on neuronal viability may be roughly comparable between the various aggregation species present, particularly oligomers and protofibrils. In contrast, fibrillar preparations of both mutants inhibited neuronal viability to a significantly greater extent than fibrillar wild type A␤-(1-42) (Fig. 5, B and D). Thus, both fibril structure and effects on neuronal viability are altered by both the E22G and E22Q mutations, producing species comparable in neurotoxicity to wild type A␤-(1-42) oligomers (comparing Fig. 5, A and C to B  and D). Although wild type A␤-(1-42) oligomers are more neurotoxic than wild type A␤-(1-42) fibrils, these data suggest that in the rare cases of these genetic mutations, fibrils may be the more toxic species. DISCUSSION A␤-(1-42) peptide can be a very challenging reagent with which to work. Its amphipathic sequence and strong tendency to self-aggregate complicates characterization of both its structure and function. Significant lot-to-lot variability affects both aggregation behavior and biological activity. We have recently taken steps to minimize this variability by removing what we refer to as "structural history." Structural history is any secondary, tertiary, or quaternary structure that can act as a template or seed driving the bulk of a peptide solution down a particular aggregation pathway. These structural seeds would not be detected during synthesis using conventional quality control methods that focus primarily on chemical purity, not structural heterogeneity. Starting with a mono-dispersed peptide preparation, we describe the development of two aggregation protocols that consistently produce extensively oligomeric or fibrillar populations of wild type A␤-(1-42) within 24 h. With  1, 2, 5, and 6) and fibrillar (lanes 3,  4, 7, and 8) preparations of A␤-(1-42) wild type (lanes 1, 3, 5, and 7), E22Q (lanes 2 and 4), and E22G (lanes 6 and 8) separated by SDS-PAGE on a 12% Nu-Page Bis-Tris gel and probed with monoclonal antibody 6E10 (recognizing residues 1-17 of A␤). Oligomeric and fibrillar preparations of A␤ were prepared as described in the legend to Fig. 1. the solubilization and aggregation protocols utilized, we obtain consistent and reproducible results for both A␤-(1-42) structure as determined by AFM and neurotoxic activity using multiple lots of peptide from both synthetic and recombinant sources.
To further investigate the relationship between peptide aggregation state and neuronal viability, we applied the A␤-(1-42) oligomer-and fibril-forming conditions to A␤-(1-42) with the Dutch E22Q or Arctic E22G mutations. These mutations induce an increase in the rate of fibril or protofibril formation in A␤- , respectively, without affecting the generation of A␤ or the A␤(1-42)/A␤(1-40) ratio and actually reduce A␤-  in transfected cells (31,38,41,42). The increase in protofibril and fibril formation of the mutants under the oligomerforming conditions did not induce a consistent change in neuronal viability. However, the increase in fibril diameter in the fibril preparations of particularly the Arctic mutation correlated with a significant decrease in neuronal viability. Thus, although wild type A␤-  oligomers are more neurotoxic than wild type A␤-(1-42) fibrils, these data suggest that in the rare cases of these genetic mutations, fibrils may be the more toxic species.
The molar concentration or the number of "units" of A␤ monomer in the unaggregated preparation is greater than oligomers, which is greater than fibrils, potentially complicating the interpretation of effects on neuronal viability on the basis of functional units. This assessment is further complicated because the cellular mechanism of A␤ induced neurotoxicity, and whether this mechanism is the same for oligomeric and fibrillar species remains unclear. In addition, it is certainly possible that the conformation of the various A␤ species was altered during the incubation period with the cells. For example, Van Nostrand and coworkers (33,40) hypothesize that it is the fibrillization of A␤ on the surface of smooth muscle cells that induces toxicity. We were not able to obtain AFM images of the three A␤ preparations at the end of the incubation period with the Neuro-2A cells because of the low peptide concentration relative to the presence of other proteins. However, the studies presented in this paper focus on the differences in A␤ assemblies that begin from starting material identical in structure and amount. The concentrations of the unaggregated, oligomeric, and fibrillar A␤-(1-42), A␤-(1-40), and A␤-(1-42) mutant preparations are based on the initial peptide mass. Thus, the differences in activities reported most likely arise from the conformational differences between the unaggregated, oligomeric, and fibril preparations. In the present study, the neuronal incubation conditions were comparable for the various A␤ preparations, suggesting that the significant and highly reproducible differences observed in neuronal viability emanated from the initial conformation of the peptide species.
Although genetic data indicate a central role for A␤ in the etiology of AD, the active form of the peptide that produces the pattern of neurodegeneration observed in the disease has not been definitively identified. The results presented herein support recent publications that suggest an oligomer/amyloid-derived diffusible ligand/protofibril form of A␤ may be responsible for the neurotoxicity that underlies AD pathology (3-7, 19 -26). Until recently, the involvement of A␤ in AD was interpreted primarily by means of the amyloid hypothesis whereby A␤ adopted a fibrillar structure that was the major component of senile plaques, and these plaques were neurotoxic. This hypothesis is further supported by the in vitro observation that fibrillar A␤ is cytotoxic (14,18). However, several clinical manifestations of AD cannot be entirely explained by the traditional amyloid hypothesis. First, there is an imperfect correlation between the number, location, and distribution of amyloid plaques and parameters of AD pathology, including the degree of dementia and neurodegeneration (51,52), although soluble A␤ concentrations in brain are highly correlated with severity of disease in AD (11,12). Indeed, total  C and D, striped bars). Oligomeric and fibrillar preparations of A␤ were prepared as described in the legend to Fig. 1. The graph represents the mean Ϯ S.E. for n Ն 8 from triplicate wells from at least two separate experiments using different A␤ preparations. *, significant difference between oligomers and fibrils (p Ͻ 0.01). levels of A␤ correlate with cognitive decline even in the absence of detectable amyloid plaques, leading Naslund et al. (53) to suggest that neurofibrillary tangles and initial neuritic changes may be caused by soluble A␤. Second, in transgenic mouse AD models, amyloid deposition in senile plaques appears after cognitive defects (54 -57). Third, in vitro inhibition of fibrillogenesis does not reduce toxicity (58 -60), whereas non-fibrillar A␤ has been shown to have toxic effects (19 -26). In addition, A␤ oligomerization is enhanced in the media of cells expressing APP or presenilin mutations, thus connecting oligomers with AD genetics (31,46). All these data challenge the exclusive relevance of A␤ fibrils in the etiology of AD while suggesting that soluble oligomers of the peptide may have pathological relevance. However, it is not our intent to suggest that there is not a correlation between specific measures of amyloid burden and dementia (61). Indeed, our own in vitro results suggest that fibrillar A␤-  with the E22Q and E22G mutations are at least as neurotoxic as wild type oligomers of A␤- .
Pathological mechanisms independent of frank protein aggregation are drawing attention in other neurodegenerative diseases, particularly those characterized by selective vulnerability. Acceleration of oligomerization, not fibrillization, is a key to ␣-synuclein mutations linked to early onset Parkinson's disease (62). In polyglutamine expansion diseases, including Huntington's disease, the number of protein aggregates can be reduced without the loss of cytotoxicity (63,64). In addition, there is a poor correlation between cell death and the number of Lewy bodies in the cortex of patients with diffuse Lewy body disease, again demonstrating a lack of correlation between neuronal damage and peptide aggregation (65). Recently, expression of mutant tau in Drosophila induced neurodegeneration without aggregation of the protein into neurofibrillary tangles (66). Thus, neurotoxicity in AD and other neurodegenerative diseases may be the result of a change in protein function independent of the formation of large aggregates.
The kinetic relationship among the various aggregation species of A␤-(1-42) and A␤-(1-40) remains unclear. For example, further investigation will be required to determine whether oligomeric A␤-(1-42) is an intermediate in fibril formation as has been hypothesized for protofibril A␤-(1-40) and A␤-(1-42) (67)(68)(69) or whether oligomers represent a separate assembly pathway. The aggregation conditions we have defined for A␤-(1-42) do not produce comparable structural or functional species of A␤- . Based on the general assumption that A␤-(1-42) follows the same assembly pathways as A␤- , albeit at an accelerated rate, this may be simply a matter of aggregation time. However, until the kinetic relationship between the various aggregation species of A␤-(1-42) and A␤-(1-40) can be resolved, it will remain difficult to determine which A␤ aggregate(s) is the toxic species potentially common to other neurodegenerative disorders with pathology dependent upon protein aggregation (7).