Identification of a mutant amyloid peptide that predominantly forms neurotoxic protofibrillar aggregates.

The amyloid peptide (Abeta), derived from the proteolytic cleavage of the amyloid precursor protein (APP) by beta- and gamma-secretases, undergoes multistage assemblies to fibrillar depositions in the Alzheimer's brains. Abeta protofibrils were previously identified as an intermediate preceding insoluble fibrils. While characterizing a synthetic Abeta variant named EV40 that has mutations in the first two amino acids (D1E/A2V), we discerned unusual aggregation profiles of this variant. In comparison of the fibrillogenesis and cellular toxicity of EV40 to the wild-type Abeta peptide (Abeta40), we found that Abeta40 formed long fibrillar aggregates while EV40 formed only protofibrillar aggregates under the same in vitro incubation conditions. Cellular toxicity assays indicated that EV40 was slightly more toxic than Abeta40 to human neuroblastoma SHEP cells, rat primary cortical, and hippocampal neurons. Like Abeta40, the neurotoxicity of the protofibrillar EV40 could be partially attributed to apoptosis since multiple caspases such as caspase-9 were activated after SHEP cells were challenged with toxic concentrations of EV40. This suggested that apoptosis-induced neuronal loss might occur before extensive depositions of long amyloid fibrils in AD brains. This study has been the first to show that a mutated Abeta peptide formed only protofibrillar species and mutations of the amyloid peptide at the N-terminal side affect the dynamic amyloid fibrillogenesis. Thus, the identification of EV40 may lead to further understanding of the structural perturbation of Abeta to its fibrillation.

The amyloid peptide (A␤), derived from the proteolytic cleavage of the amyloid precursor protein (APP) by ␤and ␥-secretases, undergoes multistage assemblies to fibrillar depositions in the Alzheimer's brains. A␤ protofibrils were previously identified as an intermediate preceding insoluble fibrils. While characterizing a synthetic A␤ variant named EV 40 that has mutations in the first two amino acids (D1E/A2V), we discerned unusual aggregation profiles of this variant. In comparison of the fibrillogenesis and cellular toxicity of EV 40 to the wild-type A␤ peptide (A␤ 40 ), we found that A␤ 40 formed long fibrillar aggregates while EV 40 formed only protofibrillar aggregates under the same in vitro incubation conditions. Cellular toxicity assays indicated that EV 40 was slightly more toxic than A␤ 40 to human neuroblastoma SHEP cells, rat primary cortical, and hippocampal neurons. Like A␤ 40 , the neurotoxicity of the protofibrillar EV 40 could be partially attributed to apoptosis since multiple caspases such as caspase-9 were activated after SHEP cells were challenged with toxic concentrations of EV 40 . This suggested that apoptosis-induced neuronal loss might occur before extensive depositions of long amyloid fibrils in AD brains. This study has been the first to show that a mutated A␤ peptide formed only protofibrillar species and mutations of the amyloid peptide at the N-terminal side affect the dynamic amyloid fibrillogenesis. Thus, the identification of EV 40 may lead to further understanding of the structural perturbation of A␤ to its fibrillation.
Alzheimer's disease (AD) 1 is the most common age-related neurodegenerative disorder. Extracellular amyloid plaques and intracellular neurofibrillary tangles are two typical pathological lesions of AD brains. Amyloid plaques, or neuritic plaques, which are more related to the AD pathogenesis, mainly consist of a cluster of heterogeneous amyloid peptides (A␤) ranging from 39 to 43 amino acids (1)(2). Among these, the A␤ with 40 amino acids (A␤ 40 ) accounts for about 90% while the less soluble C-terminally extended A␤ 42 is close to 10%. The hyperphosphorylation of tau, a microtubule binding protein, leads to the formation of neurofibrillary tangles (3)(4)(5)(6). Studies with transgenic mice show that A␤ promotes the formation of paired helical tau filaments (7)(8). Thus far, considerable experimental data favor the hypothesis that amyloid depositions in patient brains are one of the etiological factors causing AD dementia (9).
A␤ peptides are derived from consecutive processing of the amyloid precursor protein (APP) by two endopeptidases: ␤and ␥-secretases. A membrane-bound aspartyl protease, named BACE1, was simultaneously identified as the ␤-secretase (10 -13). The molecular identity of ␥-secretase has not been fully revealed yet. Nevertheless, the transmembrane protein presenilin 1 seems indisputably required for the release of the amyloid peptide from its precursor (see reviews in Ref. 14). Pathogenetic studies have manifested that the majority of mutations in APP, presenilin 1, or presenilin 2, identified from the earlyonset familial AD patients, increase either total production of A␤ or the proportion of A␤ 42 (reviewed in Ref. 15).
Monomeric A␤, when the critical concentration is reached, quickly folds into aggregated intermediate species such as oligomeric and protofibrillar forms, and finally, into insoluble fibrillar aggregates (16). Biophysical studies suggest that all these forms of A␤ are in equilibrium (17)(18). Increased production of A␤ peptides, particularly A␤ 42 , promotes amyloid fibrillogenesis and deposition in the limbic system (19 -20). In situ characterizations of brain tissues confirm the presence of aggregated A␤ fibrils (21) and SDS stable A␤ oligomers have been found in brains of Alzheimer patients (22). Earlier studies suggest that insoluble A␤ fibrils are toxic to neurons in vitro and are associated with neuronal damage in vivo (23)(24)(25)(26). Recently, studies using different approaches have demonstrated that both A␤ oligomers (27)(28) and protofibrils (29 -30) are neurotoxic as well.
During the optimization of substrates for measuring ␤-secretase activity in cells, we generated several APP mutants including APP SYEV (K670S/M671Y/D672E/A673V) and APP ISYEV (V669I/K670S/M671Y/D672E/A673V). Expression of these two constructs in cells showed increased processing of mutant APP at the ␤-secretase site (31) and produced mutated (D1E/A2V) amyloid peptides that we named EV 40 and EV 42 to distinguish them from the wild-type A␤ 40 and A␤ 42 . By examining the cell culture carefully, we found that stable cell lines expressing high levels of these two APP variants were less healthy than those expressing wild-type or Swedish APP. To understand whether the mutated residues in A␤ would change the properties of amyloid peptides and therefore cause the stress to the cell growth, we examined morphological structures and cellular toxicity of EV 40 in comparison to A␤ 40 . We found that EV 40 predominantly formed short, curvy, and sticky protofibrils that were similar to the intermediates of natural A␤ 40 fibrils. Furthermore, this form of protofibrils was slightly more toxic to cultured SHEP cells and neurons than A␤ 40 . The work reported here has been the first to show that the N-terminal sequence of A␤ affected the rate and state of amyloid fibrillogenesis. This may lead to further exploration of the kinetics of amyloid aggregation affected by cellular factors that may potentially interact with the N-terminal end of A␤.

MATERIALS AND METHODS
Peptide Synthesis-Amyloid ␤-(1-40) (A␤ 40 ) was either purchased from Polypeptide Laboratories (Torrance, CA), or together with its mutant, EV 40 (D1E/A2V), were synthesized by solid-phase methods employing an Applied Biosystems Model 433A Peptide Synthesizer. Crude peptide was dissolved in 0.05% trifluoroacetic acid, filtered and loaded onto a preparative reverse phase HPLC column (Vydac C-18, 22 ϫ 250 mm, 10 micron) with a flow rate of 4 ml/min and equilibrated with solvent A (0.1% trifluoroacetic acid in water). The column was developed with a linear gradient employing solvents A and B (0.07% trifluoroacetic acid in acetonitrile): 0 -10% B over 10 min and then 10 -50% B over 100 or 200 min, depending on each individual peptide. The column eluent was monitored by absorbance at 220 and 280 nm. Fractions were monitored on an analytical reverse phase system (Vydac C18, 4.6 ϫ 250 mm, 5 micron); solvents and conditions were as above. A linear gradient from 0 -70% B over 20 min at 1.0 ml/min was employed for this purpose. The chemical authenticity of each peptide was established by mass spectrometry employing a Micromass Platform II mass spectrometer equipped with a Hewlett Packard Series 1050 HPLC system. The identity of the peptide was confirmed by injecting 5 l of sample into the flow of 100 l/min of 1:1 methanol/water. The mass spectrometer was operated in electrospray ionization mode with needle voltage 3 kV, temperature 120°C, and cone voltage 30 V. The identity of the each peptide was confirmed by amino acid sequencing of the synthetic peptides.
Turbidity Assays of Amyloid Aggregation-Turbidity assays were carried out as previously described (32). Briefly, a 1 mM stock was made by dissolving the peptide in 0.22-m filter-sterilized 0.1% acetic acid then diluted 1:20 in calcium and magnesium-free Dulbecco's phosphate buffered saline to a final concentration of 50 M. Aliquots of the peptide solutions (250 l) were transferred to wells of Corning 96-well tissue culture plates, and the wells were tightly sealed with an adhesive sealer. The plates were constantly shaken at 800 oscillations/min on a Titer Plate Shaker (Lab-Line Instruments, Inc., Melrose Park, IL) to induce aggregation. Monitoring the aggregation process was accomplished by measuring the optical density of each well at 405 nm using a ThermoMax microplate reader (Molecular Devices, Sunnyvale, CA).
Electron Microscopy-Peptide solutions were prepared the same as for the turbidity assay to a final concentration of 50 M. Peptide solutions were incubated at 37°C for indicated times, and 5 l of each sample was then applied onto formvar-coated copper grids (200 mesh) and negatively stained with 5 l of 1% uranyl acetate for 1 min. Samples were viewed in a JEOL JEM-1230 transmission electron microscope at 80 kV accelerating voltage. Film images were captured on Kodak Electron Microscope film, and the negatives were developed and printed at an enlargement of ϫ2.5. Digital images were captured as TIFF files by a 1 megapixel Gatan Bioscan digital camera (model 792).
Circular Dichroism (CD) Spectroscopy-Peptides were prepared by dissolving them in 0.22-m filter-sterilized 0.1% acetic acid at 1 mM, the resulting solutions were then diluted 1:10 in 50 mM 0.2 m filtered sodium phosphate buffer, pH 7.2 to a final concentration of 100 M. Peptides were incubated at 37°C for indicated times and aliquots of the peptide solutions (200 l) were transferred to a quartz cell with a pathlength of 0.1 cm. Spectra were acquired using a Jasco (Easton, MD) J-715 spectrophotometer at 23°C. The CD spectra were collected from 200 -260 nm with a response of 0.25 s, scan speed of 100 nm/min, resolution of 1.0 nm and 32 cumulative scans. Secondary structures were estimated from the spectra using a principal component regression analysis method similar to one previously described (33), except ␤-turn structures were grouped together in the analysis and only 4 eigenvectors were used since the data did not extend to 178 nm. Error in the analysis of the basis spectra was within 5%.
Stock solutions of both peptides were also prepared by dissolving the peptides directly in water at a concentration of 200 M. These solutions were sampled immediately after preparation and after incubation at 37°C for 72 h. Samples (15 L) were transferred to a quartz cell with a pathlength of 0.01 cm, and CD spectra were acquired using the previous parameters except the data were collected from 184 to 260 nm and smoothed with a 13 point smoothing function. All CD spectra were converted to mean residue ellipticities for comparison.
Cell Culture-Human SHEP cells are a cell line cloned from the neuroblastoma cell line SK-N-SH that was established in 1970 from a metastatic bone tumor. SHEP cells were maintained at 37°C in a humidified, 5% CO 2 -controlled atmosphere in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum, 50 units/ml penicillin, 50 g/ml streptomycin, and 2 mM glutamine.
Preparation of Rat Primary Neurons-E18 pregnant rats were obtained from Charles River Laboratory and were anesthetized with halothane prior to cervical dislocation. Aseptically, embryos were removed and immersed in the dissociation media (Hibernate E supplemented with B27, 2 mM glutamine, and 50 units of penicillin/streptomycin). Embryo heads were cut away into a fresh 10-cm dish with dissociation medium. Overlying cartilage was dissected away, and the brains were moved to a fresh plate containing dissociation medium. Meninges and subcortical tissues were removed with fine forceps. Hippocampi and cortices were dissected and placed into 0.22 m filtersterilized 10 mg/ml papain/Hibernate solutions for 30 min at 35°C with gentle agitation. Tissue was then washed twice with warm dissociation media and dispersed into a single cell suspension by gentle trituration through a fire-polished Pasteur pipette. Cells were counted on a Coulter Counter ZM (Coulter Electronics, Luton Beds, England) and viability was determined by Trypan Blue exclusion. Cells were diluted in growth media (DMEM with 10% fetal bovine serum, 33 mM glucose, 50 units/ml penicillin, 50 g/ml streptomycin, 2 mM glutamine, B27, and 1 mM sodium pyruvate) and 1 ϫ 10 5 cells/well were plated into poly-D-lysine coated 96-well tissue culture plates. On the fourth day of culture, half of the medium was replaced with growth media supplemented with aphidicolon to a final concentration of 3.3 g/ml for 24 h after which the medium was replaced with growth media. Treatments were administered between days 12 and 14 in culture.
Cellular Toxicity Assays-Cytotoxicity was quantified by assays based on penetration of Sytox Green Nucleic acid fluorescent stains into the damaged plasma membrane (Molecular Probes, Eugene, OR). Cell culture plates were centrifuged at 250 ϫ g for 4 min. An aliquot of 50 l of medium was removed from each well and replaced with 50 l of a 2 M Sytox Green solution, prepared by diluting the 5 mM stock in Me 2 SOwith phenol red-free Opti-MEM I or phenol red-free DMEM for SHEP cells or primary neurons, respectively. The plates were incubated for 30 min at 37°C and read on a Tecan Spectrafluor Plus plate reader exciting at 485 nm and emitting at 535 nm. Alternatively, cellular toxicity was also determined by LDH assay (Promega, Madison, WI), measuring the release of the cytosolic lactate dehydrogenase at 492 nm.
Western Blot Analysis of Caspase Activation-Following amyloid challenge, cells were harvested and lysed in MAPK lysis buffer on ice (20 mM HEPES (pH 7.3), 100 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1 mM sodium orthovanadate, 0.1 mM molybdic acid, 10 mM MgCl 2 , 10 mM ␤-glycerophosphate, 5 mM p-nitrophenyl phosphate, 1 mM phenylmethylsulfonyl fluoride, 0.5% Nonidet P-40, 1 mM NaF, 5 g/ml leupeptin, 5 g/ml pepstatin, 0.05% 2-mercaptoethanol). The lysed cells were centrifuged at 10,000 ϫ g for 10 min to remove cellular debris, and nuclei and cell lysates were then quantified based on protein concentrations. Equivalent protein samples, representing ϳ7 ϫ 10 4 cells, were electrophoresed on 4 -12% Bis-Tris NuPAGE gel (Invitrogen, Carlsbad, CA). Following electrophoresis, proteins were transferred to an Immobilon-P membrane for Western analysis (Millipore, Bedford, MA). Processing/ activation of caspases was evaluated by incubating with anti-actin antibody (0.5 g/ml, Santa Cruz Biotechnology, Santa Cruz, CA) or anti-caspase-9 (1:1000 dilution, PharMingen, San Diego, CA). Following incubation with the appropriate horseradish peroxidase-conjugated secondary antibody (Santa Cruz Biotechnology), immunoreactivity was detected by chemiluminescence using SuperSignal West PICO reagent (Pierce). 40 Based on Turbidimetric Assays-A␤ fibrillogenesis is a multistep process involving nucleation, elongation, and aggregation. The kinetics of A␤ fibrillogenesis has been previously monitored by various approaches including turbidimetry (32). To gain a quick assessment of the EV 40 aggregation, we compared kinetic aggregation profiles of EV 40 to A␤ 40 by turbidimetric assays. As shown in Fig. 1, after a short duration of lag phase (ϳ40 min) under constantly controlled shaking, A␤ 40 solutions quickly became turbid. The peak of the turbidity, which reflects larger aggregation of the testing peptide, was reached at about 180-min postinitiation of shaking. Nevertheless, turbidity in the wells containing EV 40 increased at a much slower rate ( Fig.  1), suggesting that these two peptides aggregated with distinct kinetics of nucleation and oligomerization. At 360-min postinitiation of shaking, the maximum turbidity of the EV 40 -containing wells was significantly less than that of the A␤ 40 containing wells. This observation implicated that both peptides had distinct aggregation rates, and, therefore, prompted us to examine the properties of EV 40 more carefully.

EV 40 Displayed Different Aggregative Profiles from A␤
EV 40 Predominantly Formed Protofibrils While A␤ 40 Assembled to Long Fibrils-The morphology of A␤ fibrils has been well documented by the approaches of electron microscopy, atomic-force microscopy, etc. Since EV 40 had a much slower aggregation rate based on the turbidimetric assays, we suspected that the biophysical parameters of EV 40 might be different from A␤ 40 . To address this, we compared the morphological structures of these two peptides in parallel by electron microscopy. In our initial experiment, we allowed the peptides to aggregate at 37°C for 48 h followed by examining the fibrillar formation. Under these conditions, A␤ 40 formed long, rigid, and extended fibrils ϳ5-10 nm in diameter that could be longer than 500 nm in length ( Fig. 2A), consistent with the observations summarized previously (17). Under the same conditions, EV 40 did not form long and rigid fibrils as seen in the A␤ 40 preparations (comparing Fig. 2, A with B). Instead, the morphological structures found in the EV 40 preparations were short and curvy with ϳ5 nm in diameter (Fig. 2B), and ring-like structures were evident. Frequently, these short protofibrils tended to associate with each other, but did not assemble into long fibrils. In general, the morphological structure of EV 40 was similar to that of the A␤ protofibrils, identified as an intermediate that will progress to long and straight fibrils after longer incubation (17,34). The EM morphology of EV 40 also resembled the short protofibrils reported for a sample of A␤ 40 in the presence of apomorphine (35). Thus, EV 40 seemed to form an aggregated structure resembling the intermediate protofibrillar aggregates of natural A␤ 40 .
We then conducted more careful studies focusing on the dynamic protofibrillar progression of EV 40 and A␤ 40 . Without incubation at 37°C, no identifiable protofibrils or long fibrils were observed from freshly dissolved and negatively stained A␤ 40 (Fig. 2C). Occasionally, we noticed a bead-like structure in freshly dissolved A␤ 40 (similar to that seen in the background of Fig. 2A). It is unclear whether it is related to the low molecular weight (LMW) oligomers of A␤. On the contrary, EV 40 formed smaller protofibrils (Ͻ100 nm in length) even without intentional incubation at 37°C (Fig. 2D). If incubated for 4 h at 37°C, A␤ 40 proceeded to form short and irregular protofibrils (Fig. 2E), consistent with prior observations (17,30,34). Incubation of the EV 40 solution at 37°C for 4 h would allow short protofibrillar EV 40 to grow into longer protofibrils (Fig. 2F). Interestingly, even being incubated up to 8 days at 37°C, EV 40 still formed curvy protofibrillar aggregates and never formed long and straight fibrils (shown in Fig. 2G) while A␤ 40 typically formed rigid fibrils after incubation for about 48 h (Fig. 2A). It appeared that EV 40 had a very short duration of nucleation and oligomerization, but quickly locked into the protofibrillar state during prolonged incubation.
Since synthetic peptides have been shown to have slight biophysical variations in batch-to-batch preparations, we repeated electron microscopy experiments using three different batches of EV 40 to verify the above results. Again, we observed similar protofibrillar patterns using various batches of EV 40 , suggesting that the protofibrillar form of EV 40 was not due to an unusual synthesis of this peptide. A␤ 40 always produced fibrillar aggregates no matter whether A␤ 40 peptide was purchased from a commercial source or synthesized in house employing the same preparative procedures for producing EV 40 .
The Aggregative Progression Conformation of EV 40 and A␤ 40 by Circular Dichroism Spectroscopy-To determine the biophysical natures of A␤ 40 and EV 40 peptides, we employed circular dichroism to monitor the structural transition of these two peptides during aggregation. A solution of freshly prepared A␤ 40 -contained peptide in a mostly random coil and antiparallel ␤-sheet conformation as determined by principal component analysis of the CD spectrum, 47% random coil, 21% ␤-turns, 31% antiparallel ␤-sheet, Ͻ5% parallel ␤-sheet, and ␣-helical structure. Little change in the CD spectrum of the solution was detected for 2 h at room temperature but incubation at 37°C for 24 h resulted in slightly greater negative ellipticity at 215-230 nm as shown in Fig. 3A. Incubation for another 24 h produced additional spectral alterations consistent with increased antiparallel ␤-sheet.
Unlike A␤ 40 , a freshly prepared solution of EV 40 showed little conformational change for 72 h at 37°C (Fig. 3B). Interestingly, the CD spectra of EV 40 were also very similar to the spectrum of A␤ 40 collected after 24 h at 37°C. Similar conformation of EV 40 with that of a protofibrillar aggregate of A␤ 40 was consistent with the electron microscopy observations sug-  40 solutions in 96-well plates were constantly shaken to induce aggregation. The aggregation was measured by the optical density of each well at 405 nm using a ThermoMax microplate reader. gesting that EV 40 formed an aggregated structure similar to the intermediate protofibrillar aggregates of A␤ 40 .
A comparison of CD spectra for A␤ 40 and EV 40 dissolved in water at 200 M is shown in Fig. 3C. Initial spectra and spectra collected after 72 h incubation at 37°C are shown. Using a shorter pathlength than used for the previous experiments, spectral data could be collected to lower wavelengths for these solutions that did not contain acetic acid or Me 2 SO that absorb at lower wavelengths. The same trend was evident as was detected at lower peptide concentrations. A large reduction in random coil with concomitant increase in ␤-sheet conformation was detected for A␤ 40 upon incubation and although there was less random coil conformation in EV 40 initially, there was less change in conformation after incubation. EV 40 contained more ␤-sheet structure than AB 40 initially as indicated by the ratio of ellipticities at 208 compared with 215 nm. Again, this is consistent with the other data that indicates EV 40 forms protofibrils soon after dissolution but then undergoes a delay in elongated fibril formation.
EV 40 Suppresses Fibrillation of A␤ 40 -Growth of A␤ 40 fibrils can be affected by the presence of various reagents including short peptides (35)(36) and modification of residues within the A␤ region (37). To determine whether EV 40 would interfere with the fibrillar assemblies of A␤ 40 , we mixed different ratios of A␤ 40 to EV 40 in the same tube and examined the morphological structures of the mixtures. When an equal amount of EV 40 and A␤ 40 powders were mixed in a tube and dissolved using the same procedure described for the above electron microscopy experiment, we found that the morphological structure of the mixture was more similar to EV 40 preparations (Fig.  4A). The purities of each individual peptide and the mixed peptides used for the above EM experiments were verified by MALDI-TOF mass spectra (Fig. 4, B-D). This suggested that EV 40 might have suppressed fibrillation of A␤ 40 . However, this suppressive effect was not dominant since an increased proportion of A␤ 40 (e.g. 80% of A␤ 40 ) would increase the long fibrillar aggregates in the preparations (data not shown). Thus, EV 40 could interfere with the assembly of A␤ 40 into long fibrils.
EV 40 Is Slightly More Neurotoxic Than A␤ 40 -Previous experiments demonstrate that both A␤ protofibrils and fibrils induce acute electrophysiological changes and are toxic to cortical neurons (30). However, it was difficult to compare A␤ protofibrils and fibrils side by side due to limitations in the preparations of these materials. Here, we could compare the cellular toxicity of EV 40 protofibrils with A␤ 40 fibrils at the same concentrations. Prior to the treatment, we first incubated solutions of EV 40 and A␤ 40 at 37°C for 48 h to allow aggregation, and then added the aggregated peptides to cultured human neuroblastoma SHEP cells. After cells were treated for 2 days, the cell viability was examined by assays using Sytox green nucleic acid fluorescent stains (38 -39). With the increase of either EV 40 or A␤ 40 concentrations, more cells showed toxicity (Fig. 5A, ** indicates p Ͻ 0.0001 and * indicates p Ͻ 0.005 relative to control). Cells treated with A␤ 40 (50 M) had ϳ78% cell death, whereas the cells treated with the same concentration of EV 40 showed ϳ86% cell death (p ϭ 0.2, A␤ 40 versus EV 40 ).
We also compared cellular toxicity in SHEP cells treated with either freshly dissolved (without incubation as indicated with preaggregated) or aggregated (preincubation at 37°C for 48 h) peptide solutions. Cells treated with freshly dissolved EV 40 showed significantly higher cellular toxicity than A␤ 40  SHEP cells (p Ͻ 0.0001, A␤ 40 versus EV 40 ), determined both by LDH release into the medium (Fig. 5B) as well as Sytox green fluorescence (Fig. 5C). This difference was clearly consistent with the EM experiments that the non-incubated A␤ 40 presumably consists of both monomers and oligomers while EV 40 contains mainly short protofibrils (see Fig. 2, C and D).
Addition of aggregated A␤ 40 into cultured SHEP cells showed a large increase in cellular toxicity compared with non-aggregated A␤ 40 , suggesting that SHEP cells were more vulnerable to aggregated A␤ 40 than monomeric A␤ 40 . Further incubation of EV 40 only slightly increased its cellular toxicity (Fig. 5, B and C), indicating that short protofibrils of EV 40 were sufficient to kill SHEP cells. Altogether, these assays suggested that protofibrillar EV 40 exhibited modest but consistently higher toxicity than fibrillar A␤ 40 to SHEP cells in multiple independent experiments.
To confirm that EV 40 is slightly more toxic to cells than A␤ 40 , we repeated the study with cultured rat primary neurons prepared from the cortex or hippocampus. Similar to the observations in SHEP cells, we found that EV 40 treatment caused an average of 85% death of rat primary neurons (p Ͻ 0.0001) while the same concentration of fibrillar A␤ 40 resulted in 67% neuronal death on average (p Ͻ 0.001) (Fig. 5D). Hence, EV 40 did appear to be more toxic, although not substantially, to both human neuronal cells and rat primary neurons when compared with the A␤ 40 treatment.
Activation of Caspases by EV 40 Protofibrils-Neuronal toxicity induced by amyloid peptides has been partially attributed to programmed cell death (40 -41). Activation of caspase-2, -3, -6, -8, -9, and -12 in different neuronal cells by amyloid peptides have been demonstrated (42)(43)(44)(45). Although we observed cellular toxicity after SHEP cells were treated with either A␤ 40 or EV 40 , it was unclear whether this toxicity could be ascribed to apoptotic death. To explore this, we treated SHEP cells with either A␤ 40 or EV 40 in parallel and examined the activation of multiple caspases. Under our treatment conditions, we found that dramatic reductions of procaspase-9 with concomitant increases in the smaller active form were induced by either fibrillar A␤ 40 or protofibrillar EV 40 (Fig. 6), suggesting that both protofibrillar and fibrillar forms of amyloid peptides similarly activated caspases in the apoptotic pathway. We also observed similar activations of caspase-2, -3, and -8 by both A␤ 40 and EV 40 (data not shown). Likely, activation of caspase-9 occurred first and the activated caspase-9 in turn triggered sequential activation of downstream caspase-3, -2, and -8 (46). Apoptosis induced by protofibrillar amyloid species might implicate that AD neurons could be vulnerable to insult from intermediate amyloid aggregates, probably before the formation of amyloid plaques.

DISCUSSION
Assembly of amyloid fibrils occurs in multiple stages: monomers, dimers, tetramers, oligomers, protofibrils, and fibrils. Turbidimetric assays that were previously used to monitor the progression of amyloid fibrillogenesis (32, 47) may not fully reflect the authentic kinetics of A␤ fibrillogenesis. It has been shown that partial monomeric A␤ quickly folded into dimers, trimers, and oligomers during the period of preparations for the assays at either room temperature or 37°C after the peptide was freshly dissolved. Assembly of EV 40 to protofibrils occurred almost immediately after the EV 40 peptide was dissolved in 0.1% acetic acid solution followed by dilution with either phosphate-buffered saline (Fig. 2) or serum-free conditioned medium (data not shown). However, there was no significant increase in turbidity of the EV 40 solution even 6-h post-constant shaking while turbidity of the A␤ 40 solution peaked after 3 h (Fig. 1). Likely, short protofibrils should be considered to be partially soluble. Our data implicated that the increased turbidity might only reflect the formation of large aggregated amyloid fibrils while the increased formation of oligomers or protofibrils are not readily monitored with turbidimetric assays.
Thioflavin T (ThT) fluorescence assay has also been used for determining the fibril formation. The fluorescent intensity of EV 40 binding to ThT was smaller than that of A␤ 40 (data not shown), confirming that EV 40 did not form long fibrils as A␤ 40 .
Studies of ex vivo amyloid fibrils from Alzheimer's brains provided clues to the structural organization of amyloid fibrils (48). However, these ex vivo studies provided only limited resolution of fibrillar structures. Synthetic amyloid peptides aggregated in vitro have been shown to mimic the fibrillar aggregates isolated from brain tissues. Furthermore, these studies also give rise to a better understanding of the dynamic assembly of amyloid fibrils. The assembly of amyloid fibrils has been shown to be a nucleation-dependent assembly process (49 -50). This nucleation requires non-covalent association of monomeric A␤, probably involving the conversion of ␣-helical or random coiled soluble A␤ structures to ordered ␤-structures (51). The arrangement of contiguous ␤-sheet A␤ peptides eventually produces short protofibrils and insoluble ribbon-like long fibrils. Preventing formation of stable ␤-pleated structure results in inhibiting aggregation of A␤ peptides (51)(52). Our CD spectra of EV 40 indicated that EV 40 could not lead to the transition from ␣-helical or random coiled structures to ordered ␤-structures during the incubation (Fig. 3), confirming that this transition is critical for the fibril formation.
Many factors including pH, temperature, salt, and changes of residues have been shown to affect the course of A␤ fibrillogenesis (18,53). Earlier studies focusing on the C terminus of A␤ show that residues 34 -42 are critical for the seeding and subsequent evolution of amyloid aggregates (19, 32, 54 -55). The central region of A␤ also plays a critical role in the progression to mature insoluble fibrils. Residues 16 -20 were identified to serve as a binding motif during polymerization (36). Studies of peptides containing Flemish (A21G), Arctic (E22G), and Dutch (E22Q) mutations suggest that these residues greatly affect the rate or the state of fibrillar aggregations (28, 34, 45, 56 -58). For example, Flemish A␤ 40 (A21G) does not form protofibrils (34) while Arctic A␤ forms short and straight protofibrils (58). Previously, the N terminus of A␤ has been considered less crucial in affecting the course of fibrillogenesis. For example, N-terminally truncated A␤ (starting at Glu 3 , Phe 4 , Arg 5 , Gly 8 , Glu 11 , Val 12 , or Leu 17 ) does not suppress, but instead enhances in some cases, aggregation in vitro (59 -60). In this study, we found that mutation of the first two amino acids of A␤ (D1E/A2V) had a profound effect on the course of amyloid fibrillogenesis as shown in Fig. 2 (A␤ 40 versus EV 40 ). Although peptides with the first few N-terminal residues deleted still form long amyloid fibrils as shown by Tekirian et al. (61), the presence of a bulky side chain at A␤ residue 2 could potentially alter the rate of protofibrillar extension as discussed in this study. Nichols et al. (62) have suggested that growth of amyloid protofibrils can be achieved in two ways: elongation of protofibrils by monomer deposition and lateral association of protofibrils. An increase in NaCl concentration promotes protofibril- B and C, freshly dissolved peptide solutions (50 M) either without incubation (0 h incubation) or preaggregated (incubation at 37°C for 48 h) were added to cultured human SHEP cells, and this treatment was allowed to proceed for 2 days. The neurotoxicity was measured by LDH assay that is based on the release of the enzyme into the cultured media or Sytox green fluorescence assay (**, p Ͻ 0.0001; *, p Ͻ 0.001). D, rat primary cortical and hippocampal neurons were treated with preaggregated EV 40 and A␤ 40 (50 M, incubation at 37°C for 48 h) for 2 days. The viability of the primary neurons was determined by Sytox Green fluorescence assay (**, p Ͻ 0.0001; *, p Ͻ 0.001). lar extension via lateral association of protofibrils while addition of monomeric A␤ favors elongation of protofibrils by monomer deposition. We found that EV 40 forms smaller protofibrils almost immediately after the peptides were dissolved. It is likely that the concentration of monomeric EV 40 became scarce shortly after the peptide was dissolved and this might suppress the elongation of EV 40 protofibrils. Alternatively, the salt concentration was not optimal for promoting lateral association of EV 40 protofibrils. Thus, more careful studies are necessary to gain insight into the mechanism of protofibrillar extension.
Although it still remains unclear how the bulkier Glu-Val side chains impact the structure of A␤ in this region, the solubility or folding of this mutated peptide was at least altered. We found that both EV 41 and EV 42 could be synthesized by standard solid-phase methods, but could not be purified from the resin in two peptide synthesis facilities where A␤ 41 / A␤ 42 are routinely synthesized and purified to homogeneity. Apparently, C-terminal extension of even one hydrophobic residue (Val) to this mutant peptide had a profound effect on the properties of EV 40 . It is likely that EV 42 could be even more detrimental to cell growth considering the likelihood of its aggregation in the peptide synthesis resin. This could also explain our earlier observations that cells expressing high levels of APP ISYEV , liberating EV 40 and EV 42 , were unhealthy and prone to die when they reached moderately high confluency. Therefore, identification of EV 40 may implicate that side chains of the N-terminal residues may affect assembly of fibrils. An antibody that recognizes an epitope within residues 3-6 of A␤ has already been shown to arrest the formation of A␤ fibrils and to dissolve already formed fibrils (63).
Moreover, the fibrillogenesis of EV 40 apparently differs from Arctic A␤ 40 . Arctic A␤ 40 (E22G) was reported to increase the rate and quantity of protofibrillar species compared with wildtype A␤ 40 (58). Nonetheless, Arctic A␤ 40 continues to form long fibrillar species during longer incubation while EV 40 did not form long and straight fibrils during extended incubation for up to 8 days.
One of the key issues in validating the "amyloid hypothesis" is whether A␤ is detrimental to neurons. The proteinaceous components and compositions in the neuritic plaques are highly complicated. Technically, it is not possible yet to perform a neurotoxicity study by reconstituting in vivo situations using synthetic materials. Transgenic mice expressing only human APP variants fail to replicate AD pathological lesions (64). However, cellular toxicity studies using a single component (e.g. A␤ 40 or A␤ 42 ) have been extensively addressed to link A␤ fibrils to neuronal or synaptic loss seen in AD brains. A␤ in the range of 20 -100 M has been shown to be toxic to neurons (23)(24)(25)(26), but no single mechanism can account for this induced neuronal death (66). In this study, we found that EV 40 was moderately more toxic to both human neuroblastoma SHEP cells and rat primary neurons than wild type A␤ 40 . Hartley et al. (30) treated mixed brain cultures with protofibrillar A␤ 40 that was prepared from size-exclusion chromatography and found that their protofibrillar A␤ 40 was not as toxic as their long fibrillar A␤ 40 . This discrepancy could be related to the differences in initial doses of protofibrillar A␤ 40 and fibrillar A␤ 40 in their treatments. However, they demonstrate that protofibrillar A␤ 40 causes increased frequency of action potentials and membrane depolarizations (30). Here we reported that multiple caspases in SHEP cells were activated by protofibrillar EV 40 . Since the protein levels of caspase-12 in SHEP cells are below our detection limit, we treated breast carcinoma MCF-7 cells with aggregated peptides and found that both protofibrillar EV 40 and long fibrillar A␤ 40 similarly activated caspase-12 (data not shown). Altogether, these results implicated that protofibrillar A␤ 40 might play similar roles to fibrillar A␤ 40 in affecting neuronal functions such as neurotransmitter release/uptake, long term potentiation, oxidative stress and neuronal survival. Consistent with this, A␤ oligomers have been shown to disrupt synaptic plasticity and cause neuronal toxicity (27)(28). Thus, blocking initial nucleation of A␤, rather than dissolution of amyloid fibrils or inhibiting later stages of fibrillogenesis, may be a more relevant approach for AD treatment.
Thus far, clinical APP mutations at the N-terminal end of A␤ region have not been identified yet. The data from this study suggested that a D1E/A2V mutation in the A␤ peptide could be severe. Mouse genetic studies with this mutation would provide valuable insight in this aspect as well. 40 . SHEP cells were initially grown in DMEM containing 10% fetal bovine serum; then the media were replaced with serum-free Opti-MEM I (Invitrogen) while the cells were treated accordingly. The EV 40 and A␤ 40 treatments were identically prepared. As a positive control for apoptosis, one plate of SHEP cells was irradiated with UV light. A plate of SHEP cells grown in the presence of 10% serum (lane DMEM) was provided as a control to show normal levels of procaspases. The cell lysates were quantified based on protein concentrations, and normalized samples were analyzed by Western analysis. The blot was first probed with anti-caspase-9 (shown as ␣-cas-9 where ␣ represents anti, and * represents an alternative splicing procaspase-9 variant in SHEP cells), and the same blot was subsequently stripped and re-probed with anti-actin. The levels of ␣-actin (43 kDa) verified equal loading of individual samples. Serum starvation could trigger weak activation of caspase as shown in Opti-MEM I control lane.