Central to Alzheimer's disease is the misfolding of amyloid-beta (Aβ) peptide, which generates an assorted population of amorphous aggregates, oligomers and fibres. Metal ion homoeostasis is disrupted in the brains of sufferers of Alzheimer's disease and causes heightened Alzheimer's disease phenotype in animal models. In the present study, we demonstrate that substochiometric Cu2+ affects the misfolding pathway of Aβ(1–40), and the more toxic Aβ(1–42), in markedly different ways. Cu2+ accelerates Aβ(1–40) fibre formation. In contrast, for Aβ(1–42), substoichiometric levels of Cu2+ almost exclusively promote the formation of oligomeric and protofibrillar assemblies. Indeed, mature Aβ(1–42) fibres are disassembled into oligomers when Cu2+ is added. These Cu2+ stabilized oligomers of Aβ(1–42) interact with the lipid bilayer, disrupting the membrane and increasing permeability. Our investigation of Aβ(1–40)/Aβ(1–42) mixtures with Cu2+ revealed that Aβ(1–40) neither contributed to nor perturbed formation of Aβ(1–42) oligomers, although Cu2+–Aβ(1–42) does frustrate Cu2+–Aβ(1–40) fibre growth. Small amounts of Cu2+ accentuate differences in the propensity of Aβ(1–40) and Aβ(1–42) to form synaptotoxic oligomers, providing an explanation for the connection between disrupted Cu2+ homoeostasis and elevated Aβ(1–42) neurotoxicity in Alzheimer's disease.
The most common fatal neurodegenerative disease of humans is Alzheimer's disease, which directly affects greater than 26 million people worldwide . Its pathology is characterized by the self-assembly of a small peptide, amyloid-beta (Aβ), into oligomers, protofibrils and fibres [2,3]. Aβ is generated by the cleavage of amyloid-precursor protein (APP), and produces various alloforms between 38 and 43 amino acids long. Aβ(1–40) is the most abundant form. However, Aβ(1–42) aggregates more readily and demonstrates greater neurotoxicity than Aβ(1–40) [4,5]. Coupled with the observations that an increase in Aβ(1–42) relative to Aβ(1–40) is associated with early onset familial Alzheimer's disease [6–8], Aβ(1–42) is considered to be much more significant to Alzheimer's disease aetiology than Aβ(1–40).
It is generally accepted that Aβ aggregation mediates Alzheimer's disease neurotoxicity; specifically, there is considerable evidence to suggest that small diffusible oligomers of Aβ(1–42), rather than fibres are the dominant toxic form [5,9–11]. A major unanswered question remains the precise mechanism by which oligomeric Aβ exerts its toxic effect. Although controversial, one popular hypothesis involves the disruption of cell membranes at the synapse. Ultimately, Aβ might lead to thinning, puncture the lipid bilayer, or form ion channels or pores [12–16], all of which could cause synaptic dysfunction, membrane leakage and finally loss of cellular ion homoeostasis.
Alzheimer's disease is a multifactorial disease; however, what initiates Aβ entering into an amyloid cascade remains unclear [17,18]. One possible factor that can promote self-association are metal ions . Indeed metal ion homoeostasis is disrupted in Alzheimer's disease patients, with the levels of Cu2+ in the neuropil increased 4-fold; here, Cu2+ ions are found concentrated in plaques, directly bound to Aβ [20–22]. Furthermore, disrupted Cu2+ homoeostasis potentiated disease pathology in drosophila [23,24] and rabbit models  of Alzheimer's disease, and clioquinol, a Cu2+/Zn2+ chelator, has been found to reduce Aβ deposition while improving the general health of mouse models . However, these beneficial effects, have not been replicated in human trials .
Additionally, a recently identified mutation (D7H) for early-onset Alzheimer's disease has been found to grant Aβ a greater affinity for Cu2+/Zn2+, and exaggerate the effects of Zn2+/Cu2+ on Aβ aggregation . Furthermore, Alzheimer's disease pathology is influenced by prion protein interactions [29,30], which have recently been linked with copper binding to Aβ , further emphasizing the potential significance of Cu2+ in Alzheimer's disease.
A significant role for Cu2+ ions in Alzheimer's disease is supported by the observations that Aβ has a picomolar affinity for Cu2+ . As Aβ levels are also estimated to be greater than 0.5 nM at the synapse , Aβ would be expected to be able to compete with other metal chelators for the binding of Cu2+ ions, especially during depolarization, when Cu2+ is reported to reach concentrations of 15–250 μM within the synaptic cleft [34,35].
Cu2+ has been shown to bind with a full (1:1) stoichiometric complement with a similar affinity to both monomeric Aβ and mature Aβ fibres . A tetragonal complex forms between Cu2+ and Aβ's three histidine residues (His-6, His-13, His-14) along with the amino group of the N-terminus and oxygen coordination from carboxylate side chains [32,36,37]. Aβ can form a number of interchangeable Cu2+ complexes with related coordination geometries with different histidine side chains in the equatorial plane [32,38–40]. The β-pleated core of Aβ(1–40) fibres occurs between residues 14 and 40 and Cu2+ coordinating ligands fall just on the edge of this region. Solid-state NMR and pulsed EPR of the Cu2+ complex indicate that the fibrillar structure can accommodate Cu2+ coordination in Aβ(1–40) [41,42].
Potentially, metal ion coordination might influence the Aβ misfolding pathway and synaptic toxicity in a number of ways, including altering the structure of oligomer and fibre assemblies, and the kinetics of formation, as well as generation of reactive oxygen species via a redox active Cu2+ or Fe3+ . It has been known for more than a decade that binding of Cu2+ and Zn2+ is able to promote aggregation of Aβ [43,44]. However, these initial studies did not make the distinction between amorphous aggregates and neurotoxic species, and subsequent studies using the amyloid fibre specific fluorophore thioflavin T (ThT) suggested that Cu2+ only promotes formation of amorphous aggregates, and inhibits both fibre formation and cytotoxicity [45,46]. These observations seemingly conflict with animal models which state the significance of Cu2+ in AD pathology [23–26]. An explanation for this conflict has been proposed, as a study using both ThT and TEM has shown that at more physiologically relevant substoichiometric levels, Cu2+ accelerates Aβ(1–40) fibre growth, whereas suprastoichiometric levels promote amorphous aggregation and inhibit fibre growth . It was observed that different stoichiometric levels of Cu2+ also affect the ability of Aβ(1–42) to form ThT binding aggregates . However, the relationship between Cu2+ ions and Aβ misfolding is still controversial, with reports of Cu2+ preventing Aβ(1–42) fibre formation at suprastoichiometric , but also, more recently, at substoichiometric levels [49,50]. There are reports of Cu2+ promoting amorphous aggregates rather than Aβ(1–42) fibres [51–54], however, others have reported elevated Aβ(1–42) cytotoxicity with substoichiometric Cu2+ [48,50].
In the present study, we aim to obtain a more complete picture of Cu2+–Aβ(1–40)/Aβ(1–42) interactions. In the present study, we demonstrate that many seemingly conflicting observations surrounding Aβ aggregation may be attributed to Cu2+ influencing Aβ(1–40) and Aβ(1–42) very differently, as well as the significant distinction between substochiometric and suprastoichiometric Cu2+ exposure. We show that although Cu2+ accelerates Aβ(1–40) fibre formation, Cu2+ stabilizes Aβ(1–42) oligomers and protofibrils to such an extent that fibres are largely absent. These two distinct Cu2+-promoted pathways are also present for mixtures of Aβ(1–40) plus Aβ(1–42) found in vivo. Our TEM images, and a dye release assay with large unilamellar vesicles (LUVs) show marked disruption of membrane structure and permeability, compared with Cu2+ free fibres of Aβ(1–42), providing an explanation for increased Alzheimer's pathology in the presence of Cu2+ ions [23–26]. It has been suggested that distinct pathways of assembly into fibres occur for Aβ(1–40) and Aβ(1–42), and these may account for differences in their cytotoxicity . We show here that small substochiometric amounts of Cu2+ heighten the differences in assembly between Aβ(1–40) and Aβ(1–42), and for only Aβ(1–42), Cu2+ promotes almost exclusive generation of protofibrils and oligomers.
MATERIALS AND METHODS
Aβ production and solubilization
Lyophilized Aβ(1–40) was purchased commercially from Cambridge Research Biochemicals and Zinser, and lyophilized Aβ(1–42) was purchased from Imperial College London and Zinser. The peptides were synthesized using solid phase F moc chemistry, and HPLC indicated a single peak at the expected molecular mass. The peptides were also characterized using 1H NMR which confirmed that Met35 was un-oxidized. Lyophilized Aβ(1–40) and Aβ(1–42) were solubilized at a concentration of 0.7 mg/ml in water at pH 10.5, using NaOH, and then left at 5°C over 48–72 h, as described by . It is clear that solubilized preparations of Aβ have low levels of fibre nucleating assemblies as at pH 7.4 solubilized Aβ preparations typically have a lag-phase of at least 70 h when agitated, and several weeks in the absence of shaking. Aβ concentration was determined using the Tyrosine absorbance at 280 nm, ε280=1280 M−1 cm−1.
Fibre growth assay
The kinetics of amyloid formation were monitored using binding of ThT to amyloid fibres, which induces ThT to fluoresce at 487 nm; this signal is directly proportional to the amount of amyloid fibrils present. BMG-Galaxy and BMG-Omega FLUOstar fluorescence 96-well plate readers were used for the measuring of ThT fluorescence. Fluorescence readings were typically taken every 30 min, following 30 s of agitation. Fluorescence excitation and emission detection were at 440 and 490 nm, respectively. Well plates were sealed with clear polyolefin film (STARLAB) to stop evaporation.
Fibre growth kinetics are sensitive to a number of factors, including pH, concentration, agitation, ionic strength and temperature; consequently, measures were taken to reduce variance in these parameters. All fibre growth experiments were incubated at 30°C and in 30 mM HEPES buffer (because of its low affinity for Cu2+ ions) and in either 10 mM NaCl or 160 mM NaCl, for Aβ(1–42) and Aβ(1–40) respectively. The pH, a critical factor in rate of fibre growth, was adjusted to pH 7.4 with small additions of 10 mM NaOH and HCl; variation between samples was measured to be 0.05 pH units or less. Two molar equivalents of ThT was added using a fresh 2 mM stock. Copper stock solution was 10 mM CuCl2, and EDTA stock was 1 mM, and ultra-high quality (UHQ) water (10−18 Ω−1 cm−1 resistivity) was used for all experiments. The final volume in each well was 200 μl, with 96 wells per plate.
Growth curve analysis
Conversion of essentially monomeric Aβ to Aβ fibres follows a characteristic growth curve, consisting of first a lag-phase (nucleation) and then a growth phase (elongation). A growth curve can be fitted to the data, to obtain a number of empirical parameters, using the following equation :
Y represents fluorescent intensity, and x represents time. Initial fluorescence intensity is represented by υi, υf represents the final fluorescence intensity, and x0 is the time at which half maximal fluorescence is reached (t50). The apparent fibre growth rate (Kapp) is obtained by 1/τ and the lag-time (tlag) by x0–2τ. This equation allows for a slope in the initial and final parts of the growth curve, (υi + mix), (υf + mfx), rather than forcing these to be horizontal. Data were processed using KaleidaGraph 4.0 graphing and data analysis software.
Kinetic parameters have been extracted from typically six or more raw traces. A two-tailed unpaired t-test was used to confirm the significance of the difference between the kinetics with and without Cu2+ ions.
Preparation of calcein encapsulated liposomes
LUVs were created as has been previously described . A lipid mixture of phophatidylcholine (PC) (Avanti Polar Lipids Inc.), cholesterol and monosialotetrahexosylganglioside (GM1) (Avanti Polar Lipids Inc.), at a ratio of 68:30:2 by weight in a glass phial, solubilized to 5 mg/ml in a 2:1 mixture of chloroform and methanol. The mixture was then left overnight, to enable solvent to evaporate, leaving behind a lipid film. Lipid films were rehydrated to a concentration of 10 mg/ml, with 200 mM calcein, aqueous buffer at pH 7.4 with 50 mM HEPES; this mixture was left for 2 h, with vortexing every 15 min. Next, the lipid mixture was passed through an extruder once, with a 100 nm polycarbonate filter (Avanti Polar Lipids Inc.), so as to generate vesicles with a uniform range of diameters. To remove non-encapsulated calcein from the LUVs the samples were centrifuged for 4 min (at 16000 g), the supernatant discarded, and resuspended in aqueous buffer (30 mM HEPES, 160 mM NaCl, pH 7.4). This centrifugation was repeated eight times. After the final centrifugation, LUVs were resuspended to 5 mg/ml, in buffer.
Dye release membrane permeability assay
The effect of Aβ samples on membrane integrity was assessed through release of calcein from LUVs using a BMG-Omega FLUOstar fluorescence 96-well plate reader, to facilitate statisitical quantification of dye release. Calcein is largely self-quenching when encapsulated within the vesicles, giving a strong fluorescent signal upon release. Readings were taken every 30 min following 30 s of mild agitation, and temperature maintained at 30°C. Final readings were taken after 12 days. We excited samples at 485 nm, and recorded fluorescent emission at 520 nm, and 10 μM EDTA was added before final fluorescence readings were taken, to tightly chelate any Cu2+ ions which have been shown to be capable of some quenching of calcein fluorescence. LUVs were diluted to 1 mg/ml for this experiment, whereas concentration of Aβ(1–42) was 10 μM, and concentration of Cu2+ was 4 μM. Each fluorescence measurement is presented as a percentage of maximum possible calcein release, determined by the addition of the detergent Triton, to a final Triton concentration of 10 ml/l.
Aliquots of Aβ samples from the fibre growth assays, or aliquots of LUV samples, were added to flow-discharged carbon-coated 300-mesh using the droplet method, with UHQ H2O washes before and after addition of stain. Phosophotungstic acid (2%, w/v), adjusted to pH 7.4, was used to negatively stain the sample. Selected images are representative of ~30 images that were taken over 30–60 min, across at least 10 fields. Images were recorded using a JEOL JEM-1230 electron microscope operated at 80 kV, and the Olympus iTEM software package. Our TEM images of LUVs were not discernibly different to images recorded with cryoelectro-tomography , suggesting that our stain did not disrupt the LUV morphology. This was confirmed using the calcein dye release assay which showed no difference in dye release in the presence or absence of phosphotungstic acid or Cu2+ ions.
Statistical significance (P< 0.05) was determined using either a two-tailed Student's t-test, or ANOVA with Fisher's least significant difference (LSD) post-hoc test, where appropriate. Data were analysed with KaleidaGraph 4.0 graphing and data analysis software.
Copper and Aβ(1–42) oligomer assembly
We used the well-established amyloid-binding ThT fluorescence assay, as well as TEM, to examine the fibre formation kinetics and assembly of Aβ(1–42) over a range of Cu2+ concentrations. Aβ fibre growth progresses through a nucleation-dependent pathway; formation of an oligomeric nucleus precedes elongation into mature fibrils [59,60]. This classic pattern of fibre formation can be seen for Aβ(1–42) without Cu2+ present, with the lag time for nucleation being 58±3 h, and half maximal fluorescence being reached at 70±3 h (Figure 1). We found that Cu2+ inhibits Aβ(1–42) fibre formation in a concentration dependent manner. With as little as 0.1 equivalents of Cu2+, the lag time of fibre formation was increased to 89±2 h, and t50 was increased to 98±2 h (a two-tailed unpaired t-test confirms that these increases are significant with >99% confidence). For 0.4 mol equivalents of Cu2+, fibre growth was largely prevented, and for 10 equivalents of Cu2+, fibre growth was completely abolished. This experiment has been repeated a number of times, at a range of Aβ(1–42) and NaCl concentrations, with similar results. Thus, the preliminary data, showing ThT florescence in the presence of substoichiometric Cu2+ and Aβ(1–42), we previously reported as supplemental information , is not supported by the present study.
Aβ(1–42) fibre growth with Cu2+
We characterized what species of aggregate Aβ(1–42) formed with TEM (Figure 1). TEM images showed that for Aβ(1–42), in the absence of Cu2+, amyloid fibres dominate the TEM grids after 120 h incubation, with a typical straight unbranched morphology of 10–20 nm thickness (depending on the number of filaments stacked together), and typically many microns in length. However, at 0.4 equivalents of Cu2+, very few fibres were observed, with Aβ(1–42) chiefly existing as a variety of oligomeric species, ranging from 10 to 30 nm circular oligomers to ‘curly’ protofibrillar oligomers of 10–20 nm thickness, and 30–100 nm length.
A different species of aggregate is observed for 10 equivalents of Cu2+, with Aβ(1–42) primarily forming amorphous aggregates, of 100–1000 nm, with dense staining, and no repeating structure. Thus, our results show that substoichiometric levels of Cu2+ promote Aβ(1–42) oligomer formation, whereas suprastoichiometric levels favour formation of amorphous aggregates. These observations were made consistently, over numerous Cu-Aβ(1–42) preparations and TEM grids.
It is also notable that the influence of Cu2+ on Aβ(1–42) oligomer formation does not occur in a stoichiometric manner. Less than half a mole equivalent of Cu2+ relative to Aβ(1–42) is able to almost completely inhibit fibre formation, and even as little as 0.1 mol of Cu2+ causes substantial delays in fibre formation, extending the amount of time that Aβ(1–42) molecules exist as oligomers. Indeed, TEM indicates that even at just 0.1 molar equivalent of Cu2+ the majority of assemblies observed are protofibrils rather than fibres. Thus, a single Cu2+ ion can influence the assembly of a number of Aβ(1–42) molecules, promoting and stabilizing oligomer and protofibril populations over fibres.
We were next interested in whether Cu2+ was able to influence preformed Aβ(1–42) fibres. For this, we added 0.4 molar equivalents of Cu2+ to Aβ(1–42) that had been incubated for 100 h at 30°C, and for which ThT fluorescence reports completed growth of amyloid fibrils. We found that once Cu2+ was added, ThT signal rapidly dropped (within 60 min) by ~40% as shown in Figure 2. Using TEM, we found that when substochiometric Cu2+ was added to fully formed Aβ(1–42) fibres, the predominant species of aggregate formed very much resembled the oligomers generated when monomeric Aβ(1–42) is incubated with Cu2+: protofibril-like oligomers of 10–20 nm thickness, 30–100 nm length. The same observation was made when substochiometric levels of Cu2+ were added to more mature Aβ(1–42) fibres. In addition, different substochiometric levels of Cu2+ were added: 0.4, 0.6, 0.8 mol equivalents. All ratios showed a similar reduction in ThT signal, of ~40%. It is clear Cu2+ not only delays Aβ(1–42) from forming fibres, but maintains Aβ(1–42) in a protifibrillar form, and even reverts fibres back to oligomers.
Addition of Cu2+ to preformed Aβ(1–42) fibres
To assess whether the protofibril-like oligomers formed by Aβ(1–42) in the presence of Cu2+ would persist following Cu2+ removal, we exposed them to EDTA, a very tight Cu2+ chelator. Our results illustrate that upon removal of Cu2+ with EDTA, Aβ(1–42) resumes typical amyloidogenecity, rapidly forming fibres with no lag time, as determined by ThT fluorescence, shown in Figure 3. After the addition of EDTA, TEM images indicate the formation of the typical long unbranched fibres.
Addition of EDTA to Cu2+-generated Aβ(1–42) oligomers
Copper and Aβ(1–40) fibre assembly
The behaviour of Aβ(1–40) in the presence of substoichiometric amounts of Cu2+ significantly differs from Aβ(1–42). Rather than completely inhibit fibre formation, 0.4 equivalents of Cu2+ markedly accelerates Aβ(1–40) fibre formation (Figure 4). In the absence of Cu2+, Aβ(1–40) fibre growth had a lag time of 84±7 h, and a t50 of 91±7 h, and addition of just 0.4 equivalents of Cu2+ halved both lag time and t50, to 40 1 and 45±1 h, respectively. A two-tailed unpaired t-test confirms that these increases are significant with >95% confidence. In contrast, 10 molar equivalents of Cu2+ ions prevents Aβ(1–40) from forming fibres. This confirms our previous observation that for Aβ(1–40) substoichiometric levels of Cu2+ accelerate fibre formation, whereas suprastoichiometric levels of Cu2+ prevent fibre formation .
Aβ(1–40) fibre growth with Cu2+
The striking differences in behaviour of Aβ(1–40) and Aβ(1–42) in the presence of Cu2+ is also apparent from the TEM images, see Figure 4. We observed that for Aβ(1–40) without Cu2+ and for Aβ(1–40) grown with 0.4 equivalents of Cu2+, fibres predominate; there was no discernible difference in morphology for these two samples, both being 10–20 nm thick, typically >1000 nm long, and with twists at regular intervals, with periodicity ranging from 30 to 300 nm. With 10 equivalents of Cu2+, we found amorphous aggregate formation is favoured, with no regular structure, a greatly variable size, and dense staining. Many images supporting this behaviour were observed.
Effect of Cu2+ on fibre growth for mixtures of Aβ(1–42) and Aβ(1–40)
Following the observation that Cu2+ influences the misfolding pathway of the two alloforms, Aβ(1–42) and Aβ(1–40), in strikingly differing ways, we became interested in how substochiometric Cu2+ would influence mixtures of the two peptides, as is the case in vivo. We found that when we kept the total concentration of peptide at 10 μm (Aβ(1–42) + Aβ(1–40)), a higher ratio of Aβ(1–42) relative to Aβ(1–40) (in the presence of Cu2+) caused a decrease in the intensity of the final fluorescent signal, as seen in Figure 5A. This reduction appeared to be due to a decrease in the total Aβ(1–40) concentration, rather than an interaction between the two peptides. This is strongly supported by our second experiment, in which the concentration of Aβ(1–40) was kept constant, at 10 μm, as the concentration of Aβ(1–42) increased from 0 to 10 μm. As can be seen in Figure 5B, addition of Aβ(1–42) to Aβ(1–40), in the presence of Cu2+, did not decrease the final fluorescent signal, as would be expected if Aβ(1–40) formed oligomers with Aβ(1–42), nor did fluorescent signal increase, as would be expected if Aβ(1–42) formed mature fibres with Aβ(1–40). These data strongly suggest that at equilibrium (indicated by a plateau in the ThT signal) Cu2+–Aβ(1–40) and Cu2+–Aβ(1–42) are distinct, with all Aβ(1–40) forming fibres and all Aβ(1–42) forming oligomers in the presence of Cu2+. This is supported by our TEM images: as can be seen in Figure 6, in a mixture of Aβ(1–42) and Aβ(1–40), species representative of Aβ(1–40) fibres and species representative of Cu2+–Aβ(1–42) oligomers are both observed.
Aβ(1–40):Aβ(1–42) fibre growth with Cu2+
Images of Aβ(1–40):Aβ(1–42) mixtures with Cu2+
However, we note that although the two peptides appear to adopt two distinct final misfolded states in the presence of Cu2+, with Cu2+–Aβ(1–40) forming fibres and Cu2+–Aβ(1–42) forming oligomers, there is clear interaction between the two misfolding pathways. As illustrated in Figure 5C, the presence of Aβ(1–42) is able to dramatically frustrate growth of Aβ(1–40) fibres in the presence of Cu2+. Aβ(1–40) fibre growth, with 0.4 equivalents of Cu2+, had a t50 of 38±4 h, and a lag time of 27±5 h; addition of just 0.3 mol equivalents of Aβ(1–42) (which is the Aβ(1–40):Aβ(1–42) ratio found in a familial Alzheimer's disease) more than doubled t50 and lag time, to 62±2 and 50±1 h, respectively. A two-tailed unpaired t-test confirms that these increases are significant with >99% confidence.
Effect of Cu2+ on Aβ(1–42) lipid membrane structure and permeability
A popular hypothesis suggests that Aβ synaptotoxicity is the result of cellular membrane interactions and disruption. We were therefore interested in how Cu2+ binding to Aβ(1–42) might influence the interaction between Aβ(1–42) and lipid membranes, as the presence of Cu2+ facilitates the study of an almost exclusively oligomeric/protofibrillar population which is stable over long periods of time. Using TEM, we observed the interaction of liposomes with Aβ(1–42) fibres, and compared this to the effects of Cu2+-generated oligomers on the lipid bilayer, as shown in Figure 7. In the absence of Aβ(1–42), we observed smooth, spherical vesicles, of 50–250 nm in diameter. We found that in the presence of Aβ(1–42) fibres, some LUVs were distorted from their regular spherical appearance where they made contact with Aβ(1–42) fibres, particularly at their ends, as has recently been observed for amyloid fibres of β2-microglobulin . TEM of LUVs in the presence of Cu2+-generated Aβ(1–42) oligomers and protofibrils show markedly more distortions of the vesicle membrane, shown in Figure 7. Notably, the short curly morphology of the oligomers and protofibrils, penetrate the membrane and disrupt the membrane more readily, leading to much greater distortion of the LUVs than that typically observed with Aβ(1–42) fibres. In particular, there are complete breakages/discontinuations in the lipid bilayer, highlighted in Figure 7, which were not observed with Aβ(1–42) fibres. In addition, there are a greater number of oligomers observed relative to Aβ(1–42) fibres, from the same concentration of Aβ(1–42) monomers.
TEM images of LUVs in the presence of Aβ(1–42) fibres and Cu2+-generated oligomers
We used a dye release assay with LUVs to quantitatively measure the influence of Aβ(1–42) fibres and Cu2+-generated Aβ(1–42) oligomers on membrane permeability. We measured calcein dye release from LUVs, by monitoring fluorescence every 30 min. Typically, in the absence of Aβ the LUVs remained stable for more than 300 h with negligible dye release. We found that incubation of LUVs with Cu2+-generated Aβ(1–42) oligomers results in significantly greater release of calcein than when LUVs are incubated with Aβ(1–42) fibres alone, Figure 8. Furthermore, both samples of Aβ(1–42) induced significantly greater fluorescence release than control (LUVs in the absence of Aβ), as determined by a one-way ANOVA followed by Fisher's LSD post-hoc test.
Increase in liposome permeability in the presence of preformed Aβ(1–42) fibres and Cu2+-generated oligomers
In our investigations, we have found that substoichiometric levels of Cu2+ influence the misfolding pathways of Aβ(1–40) and Aβ(1–42) in markedly distinct ways. For Aβ(1–40), Cu2+ accelerates the rate of fibre formation, but had little influence on the morphology or amount of fibres generated. The mechanism by which Cu2+ increases rate of Aβ(1–40) fibre growth has been suggested to originate from the addition of the two positive charges making Aβ more neutral in overall net charge at pH 7.4, and consequently more prone to self-association . However, for Aβ(1–42), we saw that Cu2+ greatly reduced fibre number, but instead promoted the formation of oligomers and short “curly” protofibrils. These typically transient forms of Aβ were surprisingly stable in the presence of substochiometric Cu2+ over a period of weeks at 30°C. Furthermore, preformed mature fibres of Aβ(1–42) rapidly dissociated in the presence of Cu2+, also generating oligomers and protofibrils, indicating that in the presence of substochiometric Cu2+ protofibrilar and oligomeric assemblies are the thermodynamically stable form of Cu2+-loaded Aβ(1–42).
It is significant that Cu2+ is able to promote formation of Aβ(1–42) oligomers over fibres, as studies suggest that Aβ oligomers are the primary mediator of synaptotoxicity in Alzheimer's disease [5,9–11]. This is supported by the observation that soluble, diffusible forms of Aβ, monomers and small oligomers, better correlates with cognitive impairment and synaptic dysfunction than the number of larger aggregates [61,62]. The differing structures of fibres and oligomers and the extent of exposed hydrophobic residues may be key to oligomer cytotoxicity.
Cu2+ could contribute to Alzheimer's disease pathology, through binding to Aβ(1–42) and promoting oligomer formation. This is supported by our TEM images of oligomer–liposome interaction and our vesicle permeability assay, which both indicate that Cu2+-generated oligomers/protofibrils of Aβ(1–42) profoundly influence lipid membrane structure and integrity. Consequently, it appears that the increased Alzheimer's disease phenotype in animal models for which copper homoeostasis is disrupted [23–26] and in cell culture [47,48,50,63] is at least in part due to Cu2+ dramatically altering the distribution between fibre and oligomer for Aβ(1–42), and these Cu2+-generated oligomers having a more marked effect on membrane integrity than Aβ(1–42) in the absence of Cu2+. Additionally, at neutral pH, the contribution of two positive charges upon Cu2+ binding to Aβ makes Aβ more neutrally charged overall, which may facilitate greater penetration of the hydrophobic lipid-bilayer.
We were surprised that just two amino acids at the C-terminus of Aβ (Ile, Ala) that are not directly involved in Cu2+ coordination [32,38,40] have such a profound influence on the extent of oligomer formation. The marked difference in the effect Cu2+ has on the two Aβ alloforms may arise from distinct differences in the pathway to fibril assembly [55,64]. Unlike Aβ(1–40), it is believed Aβ(1–42) fibre growth proceeds through an initial stage of oligomerization, in which pentamer/hexamer units are formed that ultimately reassemble to form fibres . It may be possible that Cu2+ binding stabilizes these otherwise transient species that are more readily formed by Aβ(1–42), conceivably by cross-linked inter-molecular coordination , promoting a more thermodynamically stable oligomer rich population of Aβ(1–42). Substochiometric levels of Cu2+ are sufficient to promote almost exclusive formation of oligomers and protofibrils: our data suggest that a single Cu2+ ion is sufficient to alter the stability of a number of Aβ(1–42) molecules in an oligomeric form. Interestingly, recent studies using solid-state NMR suggest differences in secondary structure between fibres and oligomers occur in the N-terminus of Aβ(1–42) (residues 3–14) . It is these residues that adopt a β-strand structure in the oligomer structure but not in fibres. Interestingly, it is these N-terminal residues that are directly involved in Cu2+ coordination via His-6, His-13 and His-14 .
In vivo, there is a mixture of Aβ(1–40) and Aβ(1–42) present, the ratio of which increases from 9:1 to 7:3 in familial forms of Alzheimer's disease [6–8]. How these two peptides influence the misfolding of the other is a subject of intense interest. It appears Aβ(1–42) is able to frustrate the fibre formation of Aβ(1–40) [67–70]. Our observations for Cu2+-loaded mixtures of Aβ(1–40) and Aβ(1–42) provides a fascinating insight. The data suggest that in the presence of Cu2+, Aβ(1–42) can frustrate Aβ(1–40) fibre formation, prolonging the time that Aβ(1–40) spends as prefibrillar oligomers, but ultimately only Cu2+–Aβ(1–40) form fibres, whereas Aβ(1–42) forms stable protofibrils and oligomers.
Our study illustrates that Cu2+ accentuates differences in the self-assembly pathways for Aβ(1–42) and Aβ(1–40). This suggests the significance of Aβ(1–42) in Alzheimer's disease aetiology could be related to its very different aggregation properties in the presence of Cu2+ ions released at the synapse. Thus, the elevated levels of Cu2+ that occur with age  may serve as an important factor in Alzheimer's disease onset, through stabilizing neurotoxic oligomeric species of Aβ(1–42). Supporting a role for Cu2+ in Alzheimer's disease, a recent unbiased screen indicated that copper binding compounds, clioquinol in particular, ameliorated Aβ toxicity in a yeast model of Alzheimer's disease . Indeed, our investigations show that chelation of Cu2+ from Aβ(1–42) results in the rapid conversion of Aβ(1–42) oligomers to the less neurotoxic amyloid fibres.
All authors conceived the research. Christian Matheou and Nadine Younan performed the research and analysed the data. Christian Matheou and John Viles wrote the paper. All authors gave final approval of the version to be published.
The present work was supported by Wellcome Trust [grant number 093241/Z/10/Z], and a Biotechnology and Biological Sciences Research Council Quota Studentship.