A Ganglioside-induced Toxic Soluble Aβ Assembly

The mechanism underlying plaque-independent neuronal death in Alzheimer disease (AD), which is probably responsible for early cognitive decline in AD patients, remains unclarified. Here, we show that a toxic soluble Aβ assembly (TAβ) is formed in the presence of liposomes containing GM1 ganglioside more rapidly and to a greater extent from a hereditary variant-type (“Arctic”) Aβ than from wild-type Aβ.TAβ is also formed from soluble Aβ through incubation with natural neuronal membranes prepared from aged mouse brains in a GM1 ganglioside-dependent manner. An oligomer-specific antibody (anti-Oligo) significantly suppresses TAβ toxicity. Biophysical and structural analyses by atomic force microscopy and size exclusion chromatography revealed that TAβ is spherical with diameters of 10–20 nm and molecular masses of 200–300 kDa. TAβ induces neuronal death, which is abrogated by the small interfering RNA-mediated knockdown of nerve growth factor receptors, including TrkA and p75 neurotrophin receptor. Our results suggest that soluble Aβ assemblies, such as TAβ, can cause plaque-independent neuronal death that favorably occurs in nerve growth factor-dependent neurons in the cholinergic basal forebrain in AD.

The poor correlation between amyloid load in the brain and the degree of neurological deficits in patients with Alzheimer disease (AD) 2 (1) or animal models of AD (2, 3) argues against amyloid fibrils being the primary toxic A␤ species. Recently, soluble A␤ assemblies, also referred to as A␤ oligomers (4), protofibrils (5,6), or A␤-derived diffusible ligands (7), have attracted attention because of their potency to impair neuronal function or induce neuritic degeneration (7)(8)(9)(10)(11)(12)(13). Several possi-bilities have been proposed in regard to the toxicities of soluble A␤ assemblies (e.g. the binding of assemblies to target molecules on neuronal membranes (7,14) and the ubiquitous disruption of the plasma membrane in association with the perturbation of ionic homeostasis (15)). It is also noteworthy that neurotoxicities induced by soluble A␤ assemblies are mediated, at least in part, by the activation of signal transduction pathways, including those involving Src family kinases, extracellular signal-regulated kinase, or sphingomyelinases (7,11,16,17). Notably, the level of soluble A␤ assemblies increases in the brain and cerebrospinal fluid of AD patients (18,19,20,21,22), and oligomer-specific immunoreactivity is readily observed in the AD brain (23). Furthermore, the inhibition of long term potentiation and the impairment of cognitive function in vivo can be induced by natural A␤ oligomers (9,24) or a specific A␤ assembly called A␤ ૽ 56, which has recently been isolated from Tg2576 mice (expressing a human amyloid precursor protein variant-linked familial AD) (25). Additionally, recent studies using AD mouse models revealed that soluble A␤ assemblies may play a role in the induction of tau pathology (26) and that the genetic deletion of ␤-secretase, which is responsible for A␤ production, rescues temporal memory deficit in conjunction with the suppression of the increase in the levels of cerebral A␤-derived diffusible ligands (27). These lines of evidence indicate the pathological relevance of these soluble A␤ assemblies in AD development. However, it remains to be elucidated how these assemblies are formed in vivo.
Several mutations within the A␤ sequence have been reported to be responsible for the development of familial AD and hereditary cerebral amyloid angiopathy (28 -32). Among these mutations, the Arctic mutation, unlike other mutations, accelerates the development of clinical and neuropathological features indistinguishable from those of sporadic AD, although it does not increase A␤42 level or A␤42/A␤40 ratio (30). The pathological features induced by the Arctic mutation, including predominant A␤ deposition in the brain parenchyma, have also been confirmed in transgenic mice (33). Notably, A␤ bearing the Arctic mutation shows a propensity to form neurotoxic nonamyloid assemblies, including protofibrils, amyloid pores, and small nonfibrillar assemblies (13,30,34). Thus, researchers have focused on the Arctic mutation in terms of the mechanisms underlying the formation of soluble and insoluble A␤ assemblies.
In regard to the assembly of wild-type and hereditary varianttype A␤s, we have recently observed that Arctic-type A␤, unlike other hereditary variant-type A␤s (i.e. Dutch-type, Italian-type, and Flemish-type A␤s), preferably assembles in the presence of GM1 ganglioside, as does wild-type A␤ (35,36). We also reported that GM1 ganglioside level increases in synaptosomes prepared from aged, human apolipoprotein E4 knock-in mice (37). Thus, it is possible that an alteration in the expression or distribution of GM1 ganglioside is the background to the assembly and deposition of A␤ in the brain parenchyma. This possibility has been supported by findings of recent studies as follows: 1) GM1 ganglioside level increases in membrane microdomains isolated from the frontal cortex but not from the temporal cortex, reflecting earlier and later stages of AD pathology, respectively (38), and 2) GM1 ganglioside level also increases in amyloid-positive nerve terminals obtained from the AD cortex (39).
In this study, we aimed to characterize the toxicity of assemblies formed from Arctic-type A␤ in the presence of GM1 ganglioside. We found that a toxic soluble A␤ assembly (TA␤) is formed more rapidly and to a greater extent from Arctic-type A␤ in the presence of GM1 ganglioside than from wild-type A␤. Furthermore, our results suggest that TA␤ induces nerve growth factor (NGF) receptor-mediated neuronal death. Thus, we propose that soluble A␤ assemblies, such as TA␤, are responsible for plaque-independent neuronal death that favorably occurs in NGF-dependent neurons in AD.

MATERIALS AND METHODS
Preparation of Seed-free A␤ Solutions and Liposomes-Synthetic wild-type A␤ (A␤40) and Arctic-type A␤ (A␤40) (Peptide Institute, Osaka, Japan) were dissolved in 0.02% ammonia solution at 500 M. To obtain seed-free A␤ solutions, the prepared solutions were centrifuged at 540,000 ϫ g for 3 h using an Optima TL ultracentrifuge (Beckman) to remove undissolved peptides that can act as preexisting seeds. The supernatant was collected and stored in aliquots at Ϫ80°C until use. Immediately before use, the aliquots were thawed and diluted with Tris-buffered saline (150 mM NaCl and 10 mM Tris-HCl, pH 7.4). To prepare liposomes, cholesterol (Sigma), sphingomyelin (Sigma), and GM1 ganglioside (Matreya LLC) were dissolved in chloroform/methanol at a molar lipid ratio of 50:50:0, 45:45:10, 42.5:42.5:15, or 40:40:20. The mixtures were stored at Ϫ80°C until use. Immediately before use, the lipids were resuspended in Tris-buffered saline at a ganglioside concentration of 2.5 mM, and the suspension was subjected to freezing and thawing and sonication.
A␤ Incubation in the Presence of GM1 Ganglioside-A seedfree A␤ solution was incubated at 37°C and 50 M, unless otherwise indicated, in the presence or absence of GM1 ganglioside-containing liposomes, as previously reported (40). The concentration of GM1 ganglioside in the incubation mixtures was 500 M, and the molar ratio of GM1 ganglioside in the liposomes varied, as indicated in each figure.
ThT Assay-A␤ solutions were incubated in the presence of liposomes at 50 M and 37°C for various durations. The ThT fluorescence intensity of the incubation mixtures was determined using a spectrofluorophotometer (RF-5300PC) (Shimadzu Co., Kyoto, Japan). The optimum fluorescence intensity of amyloid fibrils was measured at excitation and emission wavelengths of 446 and 490 nm, respectively, with the reaction mixture (1.0 ml) containing 5 M ThT and 50 mM glycine-NaOH at pH 8.5. The fluorescence intensity was measured immediately after preparing the mixture.
LDH Release Assay-The LDH assay was performed on medium using an LDH assay toxicity kit (Promega, Madison, WI). The degree of LDH release in each sample was assessed by measuring absorbance at 490 nm using an Emax precision microplate reader (Molecular Devices Corp., Sunnyvale, CA). Background absorbances, as assessed using cell-free wells, were subtracted from the absorbances of each test sample. Absorbances measured from three wells were averaged, and the percentage degree of LDH release was calculated by dividing the absorbance measured from each test sample following treatment with 1% Triton X-100 to induce the release of intracellular LDH according to instructions provided by the manufacturer.
Electron and Atomic Force Microscopies-For electron microscopy, the samples were diluted with distilled water and spread onto carbon-coated grids. The grids were negatively stained with 2% uranyl acetate and examined under a JEM-2000EX transmission electron microscope (Tokyo, Japan) with an acceleration voltage of 100 kV. Atomic force microscopy (AFM) assessment was performed as described elsewhere (41). Briefly, the samples were dropped onto a freshly cleaved mica. After leaving them to stand for 3 min and then washing with water, the samples were assessed in a solution using a Nanoscope IIIa (Digital Instruments, Santa Barbara, CA) set in the tapping mode (42). OMCL-TR400PSA (Olympus, Japan) was used as a cantilever. The resonant frequency was ϳ9 kHz.
Size Exclusion Chromatography-The molecular mass of TA␤ was determined using a Superose 12 size exclusion column (1 ϫ 30 cm; GE Healthcare) equilibrated with phosphate-buffered saline (pH 7.4) at a flow rate of 0.5 ml/min. Thirty-five fractions were collected and analyzed by dot blotting using anti-Oligo.
Preparation of Synaptosomes-Synaptosomes were prepared as previously described (43). A hippocampus or a whole brain minus the hippocampus was homogenized in 0.32 M sucrose buffer containing 0.25 mM EDTA. The homogenate was centrifuged at 580 ϫ g for 8 min. The supernatant was centrifuged at 145,000 ϫ g for 20 min. The resulting pellet was suspended in 0.32 M sucrose buffer without EDTA and layered over Ficoll in sucrose buffer. Following centrifugation at 87,000 ϫ g for 30 min, the synaptosome-rich interface was removed and recentrifuged to remove any remaining Ficoll.

RESULTS
Toxicity of A␤ Assembly Formed from Arctic-type A␤-We treated primary neurons with seed-free wild-or Arctic-type A␤, which had been preincubated for 2 h in the absence or presence of GM1 ganglioside (10 or 20% molar ratio in the lipids composing liposomes). Unexpectedly, extensive neuronal death was observed in the culture treated with Arctic-type A␤, which had been preincubated for 2 h in the presence of GM1 ganglioside at a 10% molar ratio in liposomes (Fig. 1A). The extent of neuronal death under this condition was greater than that under any other conditions examined in this study (Fig. 1,  A and B).
To quantitatively characterize the toxic A␤ assembly, we examined its toxicity against NGF-treated PC12 cells (PC12N cells). We found that PC12N cells are also sensitive to the toxic A␤ assembly formed from Arctic-type A␤ (Fig. 1C). We performed an LDH release assay of cultures of PC12N cells under various conditions. The level of LDH released from the PC12N cells, which were treated with the toxic A␤ assembly, increased depending on A␤ dose (Fig. 1D), GM1 ganglioside dose (Fig.  1E), and the duration of the exposure of the cells to the toxic A␤ assembly (Fig. 1F). In regard to the time course of A␤ preincubation with GM1 ganglioside, the level of released LDH increased with peak value at 2 h and then decreased in conjunction with an increase in the ThT fluorescence intensity of the incubation mixtures (Fig. 1G).
The Toxic A␤ Assembly Is Soluble-Importantly, the toxicity of the A␤ incubated in the presence of GM1 ganglioside was observed exclusively in the supernatant obtained by ultracentrifuging the incubation mixture ( Fig. 2A), suggesting that the toxic A␤ assembly is soluble. To examine the possibility that a TA␤ is formed in the presence GM1 ganglioside, we performed dot blotting using an oligomer-specific antibody (anti-Oligo) (23). TA␤ in the incubation mixtures was readily recognized by anti-Oligo (Fig. 2B). The specificity of TA␤ recognition by anti-Oligo was confirmed by the finding that TA␤ toxicity was significantly neutralized by coincubating the mixtures with anti-Oligo in the cultures of PC12N cells and primary neurons (Fig.  2C). However, coincubation with a monoclonal antibody (4396C), which inhibits A␤ fibrillogenesis through binding to GM1 ganglioside-bound A␤ as a seed (40), failed to inhibit the induction of TA␤ toxicity (Fig. 2D).
TA␤ Formation from Wild-type A␤-We then examined whether TA␤ is also formed from wild-type A␤ (A␤40). We first investigated how TA␤ is formed from wild-type A␤ in the presence of liposomes containing GM1 ganglioside. Interestingly, TA␤ is favorably formed from wild-type A␤ in the presence of GM1 ganglioside at a 15% molar ratio in liposomes (Fig.  3A). TA␤ toxicity was not significant in the nanomolar range of A␤ (Fig. 3B).
Biophysical and Structural Features of TA␤-To determine the biophysical and structural features of TA␤, we performed SDS-PAGE of the incubation mixtures containing TA␤. However, no high molecular weight bands corresponding to possible A␤ assemblies were detected. Bands were observed only after cross-linking pretreatment with glutaraldehyde ( Fig. 4A), consistent with previous findings showing that soluble A␤ assemblies are probably degraded by denaturing gel electrophoresis (6) unless they are cross-linked (44,45). A morphological analysis of TA␤ by electron microscopy failed to detect any definite structure under conditions in which protofibrils, which had been prepared as previously reported (30), were readily detectable (Fig. 4B). In contrast, spherical particles with diameters of 10 -20 nm, along with rod-shaped structures, were observed by AFM in the supernatant obtained by ultracentrifuging the incubation mixtures containing TA␤ (Fig. 4C). We then determined the molecular mass of TA␤ by size exclusion chromatography, which was followed by dot blotting using anti-Oligo. The immunoreactivity was recovered as a single peak with relative molecular masses of 200 -300 kDa (Fig. 4D). The recovery of TA␤ immunoreactivity in the same fraction was also observed in the incubation mixture containing wild-type A␤ (A␤40) and GM1 ganglioside at a 15% molar ratio in liposomes (Fig. 4D). Furthermore, the collected peak showed a significant toxicity against PC12N cells (Fig. 4E).
TA␤ Formation in the Presence of Natural Neuronal Membranes-Next, we tested whether TA␤ can be formed in the presence of natural neuronal membranes. We incubated Arctic-type A␤ in the presence of synaptosomes prepared from brains of mice from three different age groups. The degree of TA␤ formation was significantly higher in the incubation mixture containing synaptosomes prepared from the hippocampus of aged (2-year-old) mouse brains than in any other incubation mixtures, including those containing synaptosomes from the hippocampus or the whole brain minus the hippocampus from younger (1-month-old and 1-year-old) mouse brains (Fig. 5A).
To determine the possibility that an alteration in the lipid composition of neuronal membranes, particularly GM1 ganglioside, underlies the acceleration of TA␤ formation, we determined the levels of GM1 ganglioside, cholesterol, and phospholipids in synaptosomes prepared from hippocampi of young (1-month-old) and aged (2-year-old) mouse brains. Notably, the GM1 ganglioside level significantly increased, whereas cholesterol level significantly decreased with age (Fig. 5B).

Putative Mechanism Underlying TA␤-induced Neuronal
Death-To characterize cell death induced by TA␤, we performed nuclear staining with a membrane-permeable dye, Hoechst 33258. PC12N cells, which were treated with incubation mixtures containing TA␤ for 12 h, showed characteristics of apoptotic changes, including retracted neurites, shrunken   A, Western blot of supernatants of incubation mixtures containing Arctic-type A␤ (A␤40), which had been incubated at 50 M and 37°C for 24 h in the absence or presence of 500 M GM1 ganglioside (the molar ratio of GM1 ganglioside in liposomes was 10%). Ten nanograms of A␤ in the incubation mixtures was subjected to SDS-PAGE (4 -20% gradient gel) with (ϩ) or without (Ϫ) cross-linking pretreatment using glutaraldehyde. The blot was reacted with 4G8. B, electron micrographs of incubation mixture containing Arctic-type A␤ preincubated to allow protofibril formation (a) or of incubation mixture containing TA␤ formed from Arctic-type A␤ (b). Typical protofibril structures were observed in a; however, no definite structures aside from liposomes were observed in b. Bar, 100 nm. C, AFM image of fraction containing TA␤ formed from Arctic-type A␤. The supernatant obtained by ultracentrifuging (540,000 ϫ g, 3 h) the incubation mixture containing TA␤ was subjected to AFM. Spherical particles along with rod-shaped structures were observed. No definite structures were observed in the supernatants of incubation mixtures containing Arctic-type A␤ alone or GM1 ganglioside alone. The amplitude range is 0.1 V. Bar, 200 nm. D, size exclusion chromatography of incubation mixtures containing A␤, which had been preincubated in the absence or presence of GM1 ganglioside, on a Superose 12 column. Elution samples from 35 fractions were dot-blotted on nitrocellulose membranes. The blot was reacted with anti-Oligo or 4G8. The immunoreactivity with anti-Oligo was recovered as a single peak with an apparent molecular mass of 200 -300 kDa. Five representative fractions are shown. Peaks 2 and 4 correspond to fractions containing TA␤ and monomeric A␤, respectively. mAU, milli-absorbance unit. E, toxicities of peaks (2 and 4) collected from incubation mixtures containing Arctic-type A␤ (shown in D) against PC12N cells. Each column indicates the average of three values Ϯ S.D.*, p Ͻ 0.0001. cell bodies, and the condensation and fragmentation of nuclei in conjunction with an increase in the level of LDH released from TA␤-treated PC12N cells (data not shown). To determine if TA␤ toxicity is mediated by NGF receptors, we first treated PC12N cells, native PC12 cells, and primary neurons with TA␤ in the presence of exogenous NGF. In these cultures, cell death was markedly prevented (Fig. 6). We then knocked down the NGF receptors, including TrkA and p75 NTR , of PC12 cells, SY5Y cells, and primary neurons using specific siRNAs. The knockdown of p75 NTR or TrkA markedly suppressed the cell death induced by TA␤ in these cultures (Fig. 7).

DISCUSSION
Here, we show that a highly toxic soluble A␤ assembly (TA␤) can be formed more rapidly and to a greater extent from Arctic-type A␤ than from wild-type A␤. Notably, TA␤ formation requires GM1 ganglioside at certain densities. TA␤ is probably formed via a pathway different from one that leads to amyloid fibril formation. Biophysical and structural analyses by AFM and size exclusion chromatography revealed that TA␤ is spherical with diameters of 10 -20 nm and molecular masses of 200 -300 kDa. The most striking feature of TA␤ is its unique toxicity. Our results suggest that TA␤ induces the NGF receptor-mediated apoptosis of cultured cells.
Accumulating evidence suggests that soluble A␤ assemblies are formed as intermediates en route to amyloid fibril formation. This scenario is mainly supported by the formation of soluble A␤ assemblies early during the incubation period in vitro, which is frequently followed by the appearance of mature fibrils (5,6,8,13). Indeed, certain inhibitors of A␤ fibrillogenesis are potent for blocking the generation of A␤ oligomers (46). In this study, TA␤ was preferably formed in the presence of GM1 ganglioside at lower densities than those required for amyloid fibril formation (36). Furthermore, a monoclonal antibody specific to a seed for amyloid fibril formation (40) failed to inhibit TA␤ formation. These results suggest that TA␤ is formed via a pathway different from a straightforward pathway leading to amyloid fibril formation, as was previously suggested in the formation of other soluble A␤ assemblies (11,12).
In this study, monomeric Arctic-type A␤ was converted to TA␤ more rapidly and to a greater extent than wild-type A␤. The propensity of Arctic-type A␤ to form toxic nonamyloid A␤ assemblies has recently attracted interest (13,30,34); however, it remains to be clarified how the assembly of Arctic-type A␤ is accelerated compared with that of wild-type A␤. We previously found that A␤ fibrillogenesis from Arctic-type A␤ is also enhanced in the presence of SDS as well as GM1 ganglioside (36). Thus, taken together with the results of this study, it is likely that the negatively charged membrane surface is a preferred environment for Arctic-type A␤ to form soluble and insoluble assemblies. A previous study suggested that the lateral distribution of GM1 ganglioside affects the spatial arrangements of the oligosaccharide chain of a molecule (47). Thus, the conformation of GM1 ganglioside may be modulated at certain densities, providing a favorable microenvironment for TA␤ formation.
Results of this study imply that GM1 ganglioside potently accelerates the formation of not only amyloid fibrils but also the soluble A␤ assembly. It has recently been reported that A␤ oligomerization is induced in the presence of lipid rafts isolated from brain tissues and cultured cells in a ganglioside-dependent manner (48). Although further studies are necessary, it may be assumed that GM1 ganglioside-rich membrane microdomains, such as lipid rafts, provide a favorable environment that facilitates the formation of soluble A␤ assemblies, including A␤ oligomers and dimers (49).
In this study, the incubation of Arctic-type A␤ with synaptosomes prepared from aged mouse brains markedly induced TA␤ formation. Furthermore, the level of GM1 ganglioside significantly increased, whereas that of cholesterol significantly decreased with age. Our observation of an age-dependent alteration in lipid composition of neuronal membranes is in agreement with the result of a recent study of cerebral cortices of AD brains (38). Taking this together with our recent observation that the level of GM1 ganglioside in synaptosomes increases not only with age but also with the expression of apolipoprotein E4 (37), it is possible that TA␤ can be formed in the brain in association with the risk factors for AD development.
It was previously reported that A␤-derived diffusible ligands potently alter NGF-mediated signaling in cultured cells (11). Moreover, many previous studies suggested that A␤ toxicities emerge through the association with p75 NTR (50 -56) (for a review, see Refs. [57][58][59]. In particular, it is noteworthy that A␤ toxicity mediated by p75 NTR depends on a death domain (60) in the cytoplasmic part of p75 NTR molecules (56). Evidence indicates the dual function of p75 NTR : one for survival and the other for death (61) (for a review, see Refs. 57 and 58). Furthermore, a previous study revealed that heteromeric TrkA-p75 NTR complexes have different functions from homo-oligomeric TrkA or p75 NTR alone (62). Notably, the knockdown of either TrkA or p75 NTR is sufficient for suppressing TA␤ toxicity. Thus, it may be assumed that the function of heteromeric TrkA-p75 NTR complexes is FIGURE 7. TA␤ toxicity mediated by NGF receptors. PC12 cells, SY5Y cells, and primary neurons, which had been treated with siRNAs against TrkA or p75 NTR , were exposed to incubation mixtures containing Arctic-type A␤ (A␤40) at a final concentration of 25 M, which had been preincubated at 50 M and 37°C for 2 h in the absence or presence of 500 M GM1 ganglioside (the molar ratio of GM1 ganglioside in liposomes was 10%). TA␤ toxicity, which was assessed by LDH release assay, was markedly suppressed by the knockdown of TrkA or p75 NTR . Decreases in TrkA and p75 NTR expression levels were confirmed by Western blotting of cell lysates using anti-TrkA and anti-p75 NTR antibodies, respectively. Each column indicates the average of three values Ϯ S.D. *, p Ͻ 0.0001. perturbed by TA␤ binding to p75 NTR or TrkA, leading to apoptosis through the activation of the death domain of p75 NTR (for a review, see Ref. 58). However, it should be noted that conflicting evidence also exists; the expression of p75 NTR protects against the toxicity of soluble A␤ assembly or extracellular A␤ (63,64). These opposite conclusions imply that the signaling pathways of p75 NTR are complicated and that the functions of p75 NTR vary depending on cell type and context (for a review, see Ref. 57).
To date, various soluble A␤ assemblies with diverse structural features have been detected in a broad range of in vitro and in vivo studies, which employed different techniques in preparing or isolating such assemblies. As previously reported (11,65), A␤ assembles into multiple alternative structures. Thus, at this point, it is difficult to determine whether TA␤ is identical to or distinct from previously identified soluble A␤ assemblies. However, on the basis of its biophysical features, including its SDS disaggregatability and unsuccessful detection on a carboncoated grid by EM, TA␤ probably differs from previously reported A␤ assemblies, particularly protofibrils, because most protofibrils appear to adsorb equally onto carbon-coated grids (65); moreover, no TA␤ is detected by EM under conditions in which protofibrils are readily detected. One interesting soluble A␤ assembly is A␤ ૽ 56 (25). A␤ ૽ 56 may be a candidate A␤ assembly responsible for plaque-independent cognitive decline in AD; however, its biophysical features, including molecular mass and marked stability in SDS-PAGE, make it distinct from TA␤.
Finally, this study indicates a novel pathological implication of soluble A␤ assemblies. It is well documented that early and severe neuronal loss in the cholinergic basal forebrain in AD is probably responsible for cognitive decline in AD patients. Previous studies suggested that cholinergic phenotype alone is unlikely to be a sufficient condition for inducing neuronal death in AD. Certain cholinergic neurons, such as those in the pontomesencephalon, are unaffected in AD (66). Notably, cholinergic neurons in the pontomesencephalon are free of NGF receptors, whereas those in the basal forebrain, which are early and severely affected in AD, have NGF receptors (67). Taken together, our results suggest that soluble A␤ assemblies, such as TA␤, are responsible for the loss of NGF-dependent neurons in the cholinergic basal forebrain in AD. A future challenge is the production of a monoclonal neutralizing antibody against TA␤ toxicity, which would provide promising therapeutic strategies, as suggested by in vitro and in vivo studies that selectively targeted A␤ oligomers (68,69).