Although soluble oligomeric and protofibrillar assemblies of Aβ‐amyloid peptide cause synaptotoxicity and potentially contribute to Alzheimer's disease (AD), the role of mature Aβ‐fibrils in the amyloid plaques remains controversial. A widely held view in the field suggests that the fibrillization reaction proceeds ‘forward’ in a near‐irreversible manner from the monomeric Aβ peptide through toxic protofibrillar intermediates, which subsequently mature into biologically inert amyloid fibrils that are found in plaques. Here, we show that natural lipids destabilize and rapidly resolubilize mature Aβ amyloid fibers. Interestingly, the equilibrium is not reversed toward monomeric Aβ but rather toward soluble amyloid protofibrils. We characterized these ‘backward’ Aβ protofibrils generated from mature Aβ fibers and compared them with previously identified ‘forward’ Aβ protofibrils obtained from the aggregation of fresh Aβ monomers. We find that backward protofibrils are biochemically and biophysically very similar to forward protofibrils: they consist of a wide range of molecular masses, are toxic to primary neurons and cause memory impairment and tau phosphorylation in mouse. In addition, they diffuse rapidly through the brain into areas relevant to AD. Our findings imply that amyloid plaques are potentially major sources of soluble toxic Aβ‐aggregates that could readily be activated by exposure to biological lipids.
Alzheimer's disease (AD) is a neurodegenerative disorder characterized by neurofibrillary tangles and amyloid plaques consisting of aggregated Aβ‐peptide (Hardy, 2002). Fifteen years ago, the ‘amyloid hypothesis’ for AD has been proposed (Selkoe, 1991; Hardy and Higgins, 1992), but the discrepancies between amyloid plaque load in the brain and cognitive impairment in the patient (Price and Morris, 1999) or mice (Games et al, 1995) have caused a lot of controversy in the field (Terry, 2001). This has led to the concept of ‘protofibrils’ (Harper et al, 1997; Walsh et al, 1997, 1999; Hartley et al, 1999), ‘annular assemblies’ (Lashuel et al, 2002; Bitan et al, 2003); ‘Aβ‐derived diffusible ligands’ (Lambert et al, 1998) or ‘soluble toxic oligomers’ (Podlisny et al, 1995, 1998; McLean et al, 1999; Walsh et al, 2002a; Glabe and Kayed, 2006). These species are intermediary forms between free soluble Aβ‐peptides and insoluble amyloid fibers and are toxic in vitro and in vivo, whereas mature Aβ‐amyloid fibers are largely inert (Aksenov et al, 1996). The molecular nature of these smaller assemblies of Aβ remains rather elusive (Hepler et al, 2006), as different sources, isolation procedures and biophysical techniques lead to different conclusions. A number of species have been observed: dimeric and trimeric Aβ‐oligomers (Podlisny et al, 1995, 1998; McLean et al, 1999; Walsh et al, 2002a), 56* kDa oligomer assemblies from transgenic mouse brains (Lesne et al, 2006) or larger structures that consist of 50 and more Aβ peptides (Lambert et al, 1998; Walsh et al, 1999; Nilsberth et al, 2001; Chong et al, 2003; Barghorn et al, 2005; Wogulis et al, 2005; Hepler et al, 2006; Haass and Selkoe, 2007).
A β‐Amyloid fibrils, on the other hand, are found in the amyloid plaques as large insoluble macromolecular assemblies characterized by a ‘cross‐β’ fiber‐like architecture. Mature fibrils are resistant to proteolytic cleavage, are unaffected by denaturant concentrations that unfold globular proteins and possess high thermostability (Booth et al, 1997; O'Nuallain et al, 2005). Amyloid fibrils from different proteins are biologically inert, although cytotoxicity appears primarily caused by soluble prefibrillar oligomers (Bucciantini et al, 2002). Taken together, these observations suggest an archetypal model for amyloid‐associated pathologies in which disease is caused by transient toxic aggregates that eventually convert to inert amyloid deposits. These deposits would then be simple remnants of the aggregation process playing no significant role in the disease process (Lomas et al, 1992). On the other hand, amyloid fibrils are not strictly irreversible but rather in a slow dynamic equilibrium with soluble peptide (Carulla et al, 2005; O'Nuallain et al, 2005). For instance, amyloid fibrils from the PI3 kinase‐SH3 domain recycle about half of their molecules over a period of weeks (Carulla et al, 2005). Inspired by the fact that disturbed lipid metabolism is increasingly considered as an important factor in AD pathogenesis, we have investigated the influence of biological lipids on the stability of amyloid fibrils of the Alzheimer's β‐peptide 1–42 (Aβ42) in relation to neurodegeneration. We focused on sphingolipids (SM) and gangliosides (GM), which are associated with amyloid deposits (Kakio et al, 2002; Devanathan et al, 2006). In addition, lipid raft domains containing cholesterol (CH), SM and GM promote Aβ aggregation and oligomer formation (Kakio et al, 2002; Yip et al, 2002; Zou et al, 2003; Gellermann et al, 2005; Kim et al, 2006). Peroxidized lipids and their derivatives such as 4‐hydroxynonenal are involved in amyloid aggregation linking oxidative stress to Aβ deposition (IrmingerFinger et al, 1999). Finally, phospholipids stabilize toxic oligomers generated from monomeric Aβ (Johansson et al, 2007).
Lipids convert inert amyloid fibrils to a highly toxic species
We measured the effect of lipids on mature amyloid fibrils of Aβ42, in terms of their toxicity to primary neurons in vitro. We incubated Aβ42 for 2 weeks at 25°C, at 1 mg ml−1 in 50 mM Tris‐HCl, pH 7.4, and 5 mM EDTA and controlled the presence of Aβ42 fibrils by electron microscopy. These mature Aβ42 fibrils were subsequently harvested by centrifugation and added to primary hippocampal mouse neurons at a final concentration of 25 μg ml−1 (5 μM). As expected, these mature fibers were largely inert, displaying only modest neurotoxicity as assayed by Neutral Red incorporation. However, an overnight incubation at room temperature of 250 μg ml−1 mature Aβ42 fibrils with a 2.5 mg ml−1 liposome suspension composed of synthetic lipid dioleyl phosphatidylcholine (DOPC) yielded a toxic emulsion when added to primary hippocampal neurons at a final concentration of 5 μM Aβ peptide fibril and 0.25 mg ml−1 lipid (Figure 1A). Importantly, neither lipid preparations (0.25 mg ml−1) nor mature amyloid fibrils (5 μM) alone were toxic (Figure 1A). The toxic Aβ42 fibril/lipid emulsion became partitioned into two phases that were separated by centrifugation for 20 min at 14 000 g. The supernatant fraction exhibited a high degree of toxicity, whereas the pellet was largely inert (Figure 1A). Similar lipid‐induced neurotoxicity of mature Aβ42 amyloid fibrils was also observed with various other synthetic and membrane lipids, including the gangliosides GM1, sphingomyelin (SM) and brain total lipid extract (BTE) from cow (Table I). Neurotoxicity of the supernatants from these preparations was evaluated in primary cultures of neurons by neutral red incorporation (Figure 1B). The supernatants induced apoptosis, as shown by Annexin V/propidium iodide staining (Figure 1C) and cleaved caspase‐3 staining (Figure 1D). Collectively these data strongly suggest that biological relevant lipids, including lipid extracts from brain, can induce the reversal of mature amyloid fibril to a soluble toxic species.
Lipids cause mature amyloid fibrils to disassemble into a protofibrillar species
To determine the mechanism by which lipids cause the conversion of inert amyloid fibrils into a toxic state, we proceeded to the biophysical characterization of fibril–lipid mixtures. As is clear from the introduction, the atypical behavior of Aβ42 in a number of biophysical techniques has led to conflicting interpretations about the nature of the toxic species (Hepler et al, 2006). For this reason, we combine a range of biophysical assays, to obtain a maximum of information on the nature of the toxic species we here obtained. Transmission electron microscopy revealed that amyloid fibrils (Figure 2A) were converted by lipids to an insoluble fraction containing fractured and highly intertwined amyloid material surrounded by short amyloid fragments (Figure 2B), whereas the supernatant contained protofibrillar structures (Figure 2C), confirming fibril destablization and resolubilization in the presence of lipids. Confocal microscopy using immunostaining with the antibody A11 that is specific for ‘soluble prefibrillar oligomers’ (Kayed et al, 2003) shows not only a granular decoration of material on the plasma membrane of primary neurons, but also significant internalization (Figure 2D) matching the behavior of prefibrillar toxic material extracted from AD brains (Chromy et al, 2003). Amyloid–lipid emulsions were further deposited under a sucrose gradient and centrifuged at 100 000 g for 1 h. We used the Aβ‐specific mAb 6E10 and the oligomer‐specific pAB A11 to detect Aβ species. Although both the top of the gradient and the pellet reacted with 6E10, only the top of the gradient reacted with A11 (Figure 2E), demonstrating that fibrils are indeed resolubilized and that the soluble fraction migrates in the same fraction as the liposomes, whereas insoluble amyloid material was pelleted. Dynamic light scattering (DLS) at a detection angle of 90° relative to the incident beam detected hydrodynamic radii between 10 and 100 μm in samples of mature fibrils (Figure 3A1). When lipids were added to the sample (Figure 3A2), the hydrodynamic radius dropped to a range between 100 nm and 1 μm, indicating significant heterogeneity. A sample of lipids alone was monodispersed with an apparent radius of roughly 200 nm (Figure 3A3). A similar size distribution is observed from light scattering measured at 173° (back scattering), excluding misinterpretations due to the angular dependence of light scattering (data not shown). Both the size distribution and heterogeneity observed by light scattering are in excellent agreement with sizes observed by electron microscopy, where flexible protofibrils are observed to curl into spheroid shapes with dimensions between 100 and 300 nm (Figure 2C). Further confirmation that the amyloid fibrils revert to a protofibrillar state (Walsh et al, 1997; Hartley et al, 1999) was obtained from spectroscopic analysis, which showed intermolecular β or cross‐β structure similar to that of mature amyloid fibrils (Figure 3B and C). Circular dichroism (CD) revealed an increase in the amplitude around 220 nm, but no significant shape change compared to the amyloid far UV spectrum, indicating an increase in soluble material in the amyloid–lipid mixtures with a similar β‐sheet content as the amyloid fibrils (Figure 3B). Fourier‐transform infrared (FTIR) spectra indicated that lipid‐induced protofibrils possess a similar intermolecular β‐extended structure as mature fibrils (corresponding to the spectral band at 1623 cm−1), but the difference FTIR spectrum revealed some degree of unfolding in the protofibrils as compared to the mature amyloid fibrils, as was apparent from the 1647 cm−1 band (Figure 3C). We proceeded to analyze our lipid‐induced protofibrils by size exclusion chromatography (SEC). When the supernatant of a lipid/fibril mixture was injected on an S75/HR10 column, a single peak at 15.8 ml was eluted (Figure 3D, green line), which immunostained with both the 6E10 and A11 antibodies (Figure 3D). Size determination from the elution volume yields an apparent molecular weight of approximately 9 kDa (dimeric Aβ). This estimation, however, is only valid for globular proteins that do not interact with the column matrix. These requirements are certainly not met here, as the analysis of the elution peak by TEM again clearly shows a heterogeneous mixture of protofibrillar oligomers with a size of 100–200 nm. To characterize the size distribution of the lipid/fibril mixture, we instead utilized an 18‐angles static light scattering (SLS) detector inline with the SEC column, which allows to infer size information directly from the angular dependence of the scattered light intensity in an absolute manner that is independent from shape or gel matrix interactions. SLS indicates a strong nonlinear angle dependence in the light scattering intensity (Figure 3E), consistent with objects larger than 100 nm, and calculated molecular weights of 80–500 kDa (between 20 and 90 monomeric units). The fact that a heterogeneous sample elutes as a focused peak is consistent with strong interactions with the gel matrix, as under these conditions the elution profile is no longer determined by the size but by the strength of the column interactions and no size separation is achieved. Taken together, the methods used here are in agreement with earlier analysis of the structure (Walsh et al, 1999; Hepler et al, 2006) and toxicity (Walsh et al, 1997, 2002b) of protofibrils and indicate that the sample cannot be defined by a single molecular mass.
‘Backward’ protofibrils obtained by destabilizing mature fibrils using lipids are identical to ‘forward’ protofibrils obtained in ageing solutions of monomeric Aβ
We proceeded to compare fibril‐derived protofibrils (further termed ‘backward’ protofibrils) with the extensively studied soluble protofibrils formed during the aggregation process of fresh Aβ (and further termed ‘forward’ protofibrils). When we incubated monomeric Aβ42 at 1 mg ml−1 in 50 mM Tris‐HCl, pH 7.4, and 1 mM EDTA at 25°C, we found, consistent with previous observations (Walsh et al, 1999; Bitan and Teplow, 2004; Klyubin et al, 2005), that neurotoxicity of forward protofibrils peaked between 24 and 48 h (Figure 1F), at which point the sample contains a mixture of protofibrils and some mature fibrils (Figure 4A). Subsequently, the sample becomes inert and highly enriched of mature amyloid fibrils (Figure 2A shows 2‐week‐old samples) and does not contain detectable amounts of soluble protofibrils anymore. Interestingly, toxicity of forward protofibrils and lipid‐induced backward protofibrils is very similar (Figure 4B). Soluble fractions of forward protofibrils preparations elute as a single peak at 16.9 ml on an S75/HR10 column (Figure 3D, blue line) and appear as typical protofibrils of 100–200 nm diameter by transmission electron microscopy (Figure 4A). This slight difference in elution volume between the forward protofibrils and backward, lipid‐induced protofibrils (15.8 ml) is consistent with an increase in molecular weight due to lipid association. In agreement, upon addition of 0.1% SDS (submicellar concentration) to amyloid–lipid mixtures before injection on the column, a lipid peak eluted in the void volume, whereas the lipid‐induced protofibrils now elute at the same elution volume as forward protofibrils (16.9 ml; Figure 3D). These data suggest that, under the incubation condition used here, lipid‐induced backward protofibrils have very similar morphological and biochemical properties as the protofibrils formed during the aggregation process of fresh monomers.
In vivo toxicity of backward protofibrils in the brain of adult mice
To further characterize the physiological relevance of the lipid‐induced ‘backward’ oligomers, we proceeded to bilateral intraventricular injection (3 μl) of mature Aβ42 amyloid fibrils, lipids and supernatants from amyloid–lipid mixtures into the brain of adult mice. Immunostaining of brain samples with the 6E10 antibody demonstrated the effective delivery of Aβ in the ventricles when both fibrils and backward oligomers were injected into the animals (Figure 5A). However, within 90 min of injection, we also observed Aβ42 immunoreactivity away from the needle tracks into the cortex (Figure 5A), with oligomeric preparations but not with Aβ fibrils. With the latter, staining remained mostly associated with the ventricular walls (Figure 5A). When we analyzed brain areas further away from the needle track, we observed very little immunostaining after Aβ fibril injections, whereas backward oligomers caused significant staining of the hippocampal areas (Figure 5B), indicating rapid diffusion of (some of) these species into the brain parenchym (Figure 5B). Interestingly, the single injection with backward oligomers caused also mild neurotoxicity restricted to the area of the frontal cortex (but not hippocampus) that was exposed to the highest concentrations of toxic oligomers as revealed by immunostaining for cleaved caspase 3 (Figure 5C) and phosphorylated tau (Figure 5D). Lipid or mature fibrils injection alone did not cause toxicity in the frontal cortex (Figure 5C and D).
Backward and forward protofibrils cause similar memory defects in mouse
To further assess if the lipid‐induced ‘backward’ protofibrils might cause functional deficits relevant to AD, we evaluated the acute biological effects of the injected backward oligomers in exploratory and memory/learning tests. Open‐field recording revealed increased path length covered, higher velocity and increased frequencies of center visits, but declined time spent in the center by animals injected with backward oligomers, in sharp contrast to animals injected with lipid samples or mature fibrils alone. As all animals were treated identically, the backward oligomers appear to specifically affect brain functions that cause hyperlocomotion and hyperactivation (Figure 6A). In a light–dark step through task (passive avoidance test), animals were allowed to memorize the electrical shock that follows entrance to a dark compartment. When the test was repeated 24 h later, animals injected with lipids or mature fibrils correctly recalled the electroshock and avoided to enter the dark room. Injection of backward oligomers before the shock event prevented animals from successful memory formation, as was evidenced by uninhibited entering of the dark compartment 24 h after electroshock (Figure 6B). In addition, contextual and auditory‐cue fear conditioning showed that injection of backward oligomers 90 min before conditioning severely disturbed typical freezing behavior 24 h later when the animals were exposed again to the same contextual or auditory stimulus, in contrast to the control mice injected with lipids or mature amyloid alone (Figure 6C). One week after oligomer injection, mice appeared to have recovered completely and could no longer be distinguished in open‐field activity from control or untreated animals (not shown in figure). Thus, a single injection of lipid‐induced backward oligomers appears not to have caused persistent functional defects, which is in excellent agreement with earlier studies of ‘forward’ oligomers, which were reported to have immediate but transient effects on synaptic function (Wang and Hecht, 2002; Chromy et al, 2003; Gong et al, 2003). Therefore, to directly compare the behavioral effects of forward and backwards protofibrils, we performed identical behavioral tests on animals injected with forward oligomers generated from Aβ42 monomers that were incubated for 48 h at room temperature at a concentration of 1 mg ml−1. These animals showed very similar behavioral defects: open‐field recording indicated hyperlocomotion and hyperactivation effects (result not shown) and light–dark step through tests (passive avoidance) showed that animals injected with forward oligomers failed to memorize the electrical shock, in contrast to control individuals injected with buffer (Figure 6D). In addition, contextual and auditory‐cue fear conditioning showed that injection of forward oligomers 90 min before conditioning severely disturbed freezing behavior 24 h later, in contrast to the control mice (Figure 6E).
Our results confirm the behavioral effects of forward protofibrils generated from Aβ monomers (Hartley et al, 1999; Walsh et al, 2002a; Kamenetz et al, 2003; Cleary et al, 2005; Klyubin et al, 2005; Lesne et al, 2006; Townsend et al, 2006) and show that lipid‐induced protofibrils generated from mature Aβ fibrils have very similar pathophysiological effects.
We demonstrate in the current work that amyloid fibrils, usually considered as highly stable and biologically inert structures, can be destabilized and easily reverted to soluble and highly toxic Aβ aggregates by biological lipids that are present in the brain. This suggests that part of the critical balance between toxic and inert Aβ pools is determined by the relative amounts of lipids in the direct environment of the plaques. Remarkably, the toxic species we identify shares many properties with previously characterized ‘forward’ oligomeric aggregates in terms of biophysical, cell biological and behavioral assays. Although in SEC and other biophysical assays the impression could arise that these structures are homogenous, further extensive biophysical characterization indicates that the size distribution of these oligomeric aggregates is rather heterogeneous in nature, ranging from 80 to 500 kDa, although their morphology is protofibrillar. The appearance of these species indeed varies somewhat with the assay used, likely reflecting their dynamic nature. In that regard, our findings are very similar to the overall picture emerging from the literature and synthesized recently by Hepler et al (2006) that also the classical forward oligomeric Aβ structures have to be considered as a spectrum of dynamic structures that are likely in fast equilibrium with each other. Lipids are apparently promoting the equilibrium toward the protofibrillar pools, inducing toxicity of the amyloid mixture. Interestingly our data also confirm that these oligomeric structures diffuse very rapidly throughout the brain and preferentially localize to specific regions of the brain, such as the hippocampus.
Our findings have important implications for the understanding and the treatment of amyloidoses and, in particular, AD. They suggest the possibility that inert amyloid plaques or fibrils could be turned into highly toxic oligomers when local physicochemical parameters are altered due, for instance, to a change in lipid metabolism. Our results could explain why the amount of amyloid deposits and the severity of associated disease symptoms in AD do not necessarily correlate. Individual and temporal differences in brain lipid content could indeed explain why some patients with large amounts of amyloid deposits display little symptomatic disease, although others are severely affected. In any event, our data strongly suggest that amyloid plaques, although apparently biologically inert, should not be considered as inert remnants of the aggregation process, as the amyloid fibrils they contain can, under certain conditions, be rapidly reverted to toxic species. In that sense, the amyloid plaques should rather be considered as reservoirs of toxicity.
Materials and methods
Alzheimer's β‐peptide 1–42 was purchased from Sigma‐Aldrich. All purified and synthetic lipids were obtained from Avanti Lipids (USA). Uranyl acetate was obtained from BDH.
Preparation of lipid vesicles and liposomes
All lipids were obtained from Avanti Polar Lipids (USA) except the ganglioside GM1, which was obtained from Larodan Chemicals (Sweden). The stock concentration was 20 mg ml−1 in chloroform. The various lipid mixtures discussed in the paper were prepared in Corex round‐bottomed glass tubes, dried under a gentle N2 stream and resuspended in 400 μl of diethylether for 10 min at room temperature, after which they were quickly dried in a waterbath at 50°C. The resulting film was placed under vacuum for 1 h to remove trace solvent and rehydrated in 800 μl of 50 mM Tris, pH 7.5, 1 mM EDTA and 0.1 mM NaCl. The resulting vesicle suspension was allowed to stabilize for 1 h at room temperature, sonicated for 15 s (Branson sonifier) and extruded 11 times with an Avanti mini‐extruder (Avanti Polar Lipids, USA). This suspension was purified on an S75 gel filtration column using an Akta system from GEHealthcare (UK). The approximate lipid concentration in the stock preparation was 10 mg ml−1. To obtain stable liposomes, it is not possible to use pure preparations of certain lipids such as cholesterol or GM1. Therefore, we used a more complex composition as follows: (i) pure DOPC; (ii) 50% DOPC, 50% DMPG; (iii) 35% DOPC, 35% DMPG, 30% cholesterol; (iv) 30% DOPC, 30% DMPG, 30% cholesterol, 10% GM1; (v) 30% DOPC, 30% DMPG, 30% cholesterol, 10% SM; (vi) 30% DOPC, 30% DMPG, 30% cholesterol, 10% brain total extract.
Preparation of amyloid fibrils, amyloid/lipid mixtures and Aβ oligomers
Amyloid fibrils of the Alzheimer's β‐peptide 1–42 were obtained by incubation of 200 μM peptide solution in 50 mM Tris, pH 7.5, for 2 weeks at room temperature. Amyloid fibril/lipid mixtures were prepared by diluting fibril and liposome stock solutions ¼ in liposomes buffer and incubating for 1–16 h at room temperature, shaking at 700 r.p.m. Oligomers were generated from monomers by incubation of 200 mM peptide solution in 50 mM Tris, pH 7.5, for 48 h and subsequent extensive centrifugation, and the supernatant as the oligomer‐enriched preparation was used immediately. The presence of oligomers in the supernatant was validated by probing with anti‐oligomer‐specific antibody (A11).
Additional methods can be found in Supplementary data.
Supplementary data are available at The EMBO Journal Online (http://www.embojournal.org).
ICM was supported by a PhD scholarship from the Boerhinger Ingelheim Fonds (BIF, Germany), Foundation for Basic Research in Biomedicine, and subsequently a PhD scholarship from the Fundação para a Ciência e Tecnologia (FCT, Portugal), from the Portuguese Ministry for Science and Technology (MCTES). Research in BDS laboratory is supported by a Freedom to Discover grant from Bristol Myers Squib, a Pioneer award from the Alzheimer's Association, the Fund for Scientific Research, Flanders; the Artificial SynApse (IWT‐ASAP), KU Leuven (GOA) IUAP P6/43 and a Methusalem grant of the Flemish Government. The VIB Switch laboratory was supported by a grant from the Federal Office for Scientific Affairs, Belgium (IUAP P6/43), and the Fund for Scientific Research, Flanders.
This is an open access article under the terms of the Creative Commons Attribution‐NonCommercial‐NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non‐commercial and no modifications or adaptations are made.
- Copyright © 2008 European Molecular Biology Organization