Substoichiometric Levels of Cu2+ Ions Accelerate the Kinetics of Fiber Formation and Promote Cell Toxicity of Amyloid-β from Alzheimer Disease*

A role for Cu2+ ions in Alzheimer disease is often disputed, as it is believed that Cu2+ ions only promote nontoxic amorphous aggregates of amyloid-β (Aβ). In contrast with currently held opinion, we show that the presence of substoichiometric levels of Cu2+ ions in fact doubles the rate of production of amyloid fibers, accelerating both the nucleation and elongation of fiber formation. We suggest that binding of Cu2+ ions at a physiological pH causes Aβ to approach its isoelectric point, thus inducing self-association and fiber formation. We further show that Cu2+ ions bound to Aβ are consistently more toxic to neuronal cells than Aβ in the absence of Cu2+ ions, whereas Cu2+ ions in the absence of Aβ are not cytotoxic. The degree of Cu-Aβ cytotoxicity correlates with the levels of Cu2+ ions that accelerate fiber formation. We note the effect appears to be specific for Cu2+ ions as Zn2+ ions inhibit the formation of fibers. An active role for Cu2+ ions in accelerating fiber formation and promoting cell death suggests impaired copper homeostasis may be a risk factor in Alzheimer disease.

Alzheimer disease (AD) 2 is characterized by extracellular amyloid plaques, composed predominantly of fibrillar amyloid-␤ peptide (A␤), a 39 -43-residue peptide. Genetic alterations underlying familial AD are associated with mutations or increased production of A␤, indicating that A␤ plays a central role in the disease (1).
A notable characteristic of AD is altered metal ion concentrations in the brain and disrupted metal ion homeostasis (2). Cu 2ϩ ions are found concentrated within senile plaques of AD patients directly bound to A␤ (3)(4)(5). Recent in vivo studies using a Drosophila model of AD have shown that impaired copper homeostasis enhances the toxic effects of A␤ (6). Furthermore, copper in a cholesterol high diet induces amyloid plaques and learning deficits in a rabbit model of AD (7). Other in vivo studies have shown that copper homeostasis can influence AD pathology. In contrast to the Drosophila model, transgenic mice have shown a reduced AD pathology with increased intracellular copper levels (8 -10).
Although studies of A␤ neurotoxicity suggest that small diffusible oligomers, rather than mature amyloid fibers, are the more toxic form (11,12), there remains strong evidence suggesting that amyloid plaques, or possibly intermediates of the fibrils, are critical in neuronal toxicity (13,14). A␤ oligomers may be precursors of fiber formation and may also arise from fiber fragmentation. Alternatively, oligomers may be in competition with fiber formation. Both possibilities require the self-association of monomeric A␤, and thus factors that affect fibrillization will also influence oligomer generation.
The mechanism by which A␤ is toxic is hotly debated (11,15). It has been proposed that A␤ can form ion channels or pores or can thin the membrane, all of which will cause membrane leakage and loss of cellular Ca 2ϩ ion homeostasis. One popular hypothesis is that the membrane integrity is compromised by lipid peroxidation from reactive oxygen species, which is a key feature of the pathogenesis of AD (16,17). It is well established that hydrogen peroxide mediates A␤ toxicity and the antioxidant enzyme catalase protects cells from A␤ toxicity (18 -20). A likely source of extracellular H 2 O 2 is from the Fenton redox cycling of copper or iron ions (17). We and others have shown that Cu 2ϩ bound to A␤ will readily generate hydroxyl radicals and H 2 O 2 in the presence of a physiological reductant such as ascorbate (19,(21)(22)(23). Indeed, transfer of Cu 2ϩ from A␤ to the redox-inactive metallothionein III removes A␤ toxic properties (24).
The three histidine residues within A␤ peptide form a tetragonal complex with Cu 2ϩ ions (25-35; for review, see 36,37). Recent studies point to a dynamic Cu 2ϩ complex involving imidazole coordination in both the axial and equatorial plain (25,27,35). A full (1:1) stoichiometric complement will bind to both monomeric and mature A␤ fibers with identical coordination geometry and affinity (25). Affinity measurements of the Cu 2ϩ -A␤ complex have been revised, indicating a considerably tighter affinity than previously believed, setting the conditional dissociation constant, pH 7.4, at 60 ϫ 10 Ϫ12 M (25). Extracellular monomeric A␤ levels are thought to be 5 nM (38), whereas A␤ levels are higher in plaques and at the synapse. Furthermore, extracellular Cu 2ϩ levels in the brain interstitial fluid are 100 nM. A picomolar affinity for Cu 2ϩ allows A␤ to compete for Cu 2ϩ ions with other extracellular Cu 2ϩ chelators, especially at the synapse during neuronal de-polarization where fluxes of Cu 2ϩ are reported to be 20 -250 M (39).
Studies showed more than a decade ago that Zn 2ϩ and Cu 2ϩ ions cause marked aggregation of A␤ (40,41). These initial studies did not make the distinction between amorphous aggregates, which are nontoxic to cells, and the formation of amyloid fibers. Further investigations using the fiber specific fluorophore thioflavin T (ThT) suggested that Zn 2ϩ and Cu 2ϩ only promote amorphous aggregation of A␤ and actually inhibit fiber formation and cell toxicity (42)(43)(44)(45)(46). We became interested in the factors that promote self-association of A␤, the relationship between amorphous aggregation and amyloid fiber formation, and a role for Cu 2ϩ ions in promoting fiber formation. Furthermore, we wanted to establish whether there was a link between the influence of Cu 2ϩ ions on fiber formation and the effect of Cu 2ϩ ions on cell toxicity of A␤.

EXPERIMENTAL PROCEDURES
A␤ Production and Solubilization-A␤  and A␤  were synthesized using solid phase Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry; the peptides were purchased commercially from ABC-London and Zinisser Analytic. HPLC indicated a single peak with the expected molecular mass. The peptides were also characterized by 1 H NMR, and Met 35 was confirmed to be unoxidized. Lyophilized A␤  and A␤  were solubilized by dissolving 0.8 mg/ml A␤ in water at pH 10 and then placed at 5°C for 72 h. It is clear that A␤ is essentially seed-free as A␤ preparations at pH 7.4 have a lag phase of typically 100 h. The concentration of A␤ was determined using the tyrosine absorbance at 280 nm, ⑀ 280 ϭ 1280 M Ϫ1 cm Ϫ1 .
Fiber Growth Assay-The binding of ThT to amyloid fibers was used to monitor the kinetics of amyloid formation. ThT binding to amyloids induces the ThT to fluoresce at 487 nm. This fluorescence signal is related directly to the amount of amyloid. A BMG-Galaxy fluoro-star fluorescence 96-well plate reader was used for the ThT measurements. The central 60 wells were used, whereas wells around the edge contained buffer only, to minimize evaporation effects. Readings were taken every 30 min. The well plates were subjected to 30 s of agitation prior to each fluorescence measurement. Fluorescence excitation was at 440 nm and emission detected at 490 nm.
Fiber growth kinetics are very sensitive to a number of factors that must be carefully controlled. They include the pH, concentration, agitation, temperature, and ionic strength. Solubilization of A␤ into a seed-free form is also important. Fiber growth experiments were incubated at 30°C in 160 mM NaCl. In addition, 50 mM HEPES buffer was used throughout; HEPES was used for its low affinity for Cu 2ϩ and Zn 2ϩ ions. Small adjustments were made with 10 mM NaOH or HCl to the stock A␤ solutions. The pH, a critical parameter in fiber growth rates, was measured before and after each fiber growth experiment; variations were 0.05, or fewer, pH units over the course of the experiment. Metal stocks solutions were 25 mM CuCl 2 and 20 mM ZnCl 2 or as a Cu(Gly) 2 chelate. UHQ water (10 Ϫ18 ⍀ Ϫ1 cm Ϫ1 resistivity) was used at all times.
Growth Curve Analysis-Conversion of essentially monomeric A␤ to fibrillar A␤ follows a characteristic growth curve, typically described as the lag phase (nucleation) and a growth phase (elongation). A number of empirical parameters can be obtained from the fiber growth curve, including the time needed to reach half-maximal ThT intensity (t 50 ) and the lag time (t lag ) (see Table 1 and supplemental Table S1. The t 50 is influenced by both the nucleation and elongation phases. These values can readily be extracted from the data by fitting the growth curve to the following equation (25), where Y is the fluorescence intensity, x is the time, X 0 is the time at half-height of fluorescence (t 50 ). The apparent fiber growth rate is K app ϭ 1/, and the lag time (t lag ) is X 0 Ϫ 2. This equation allows for a slope in the initial and final parts of the growth curve, (y i ϩ m i x), ( f ϩ m f x), rather than forcing these to be horizontal. The fibril growth curves have also been fitted using an alternative equation to extract a rate of nucleation and elongation (47) shown in supplemental Table S1.
Kinetics parameters have been extracted from between six and nine raw traces. Mean values with 1 S.E. are given in Table 1 and supplemental Table S1. A two-tailed unpaired t test was used to confirm the significance of the difference between the kinetics with and without a Cu 2ϩ ion.
Transmission Electron Microscopy (TEM)-A␤ samples were freeze-dried and resuspended to obtain a peptide concentration of 0.5 mg/ml. The samples were added to 200mesh carbon-coated copper grids via the droplet method, and 2% uranyl acetate was used to negatively stain the samples. Images were collected with a JEOL JEM-2010 microscope operating at 200 kV.
Cell Viability-PC12 cells were used to assess the cytotoxic effect of different A␤ preparations (48). Cells were spun down at 95 ϫ g for 5 min and resuspended in 1 ml of Opti-MEM. Opti-MEM was used due to its low protein concentration (15 g/ml total protein concentration) to minimize the presence of potential competing copper chelators. A 10-l aliquot of the cells was added to 10 l of 3 mg/ml trypan blue, and the cells were counted. The cell stock was then diluted in Opti-MEM and added to the wells in a 96-well plate to give a typi-  Fig. 1, with the S.E. shown in parentheses. A, B, and C indicate data obtained in separate experiments. Equation 1 determines the kinetics parameters: t (50) , (X o ); t (lag) , (X 0 Ϫ 2dx) and; k app , (1/dx) with the number of traces shown in n. 1 H NMR-The pH-dependent protonation state of the three histidine side chains of A␤ were determined from 1 H NMR chemical shifts of the ⑀H and ␦H protons, recorded at 0.5-pH unit intervals between pH 5 and 9, as shown in supplemental Fig. S7. The singlet chemical shifts of the ⑀ and ␦ protons were readily identified from the one-dimensional 1 H NMR spectra. His C⑀H and C␦H assignments were confirmed using two-dimensional 1 H TOCSY spectra, using standard acquisition parameters and a spin-lock of 60 ms. Assignment of the His 6 , His 13 , and His 14 residues was based on previous 1 H NMR studies with H6A, H13A, and H14A analogs (49). The A␤(1-28) fragment was used rather than full-length A␤(1-40) to improve solubility. 10 M 4,4-dimethyl-4-silapentane-1-sulfonic acid was used as a reference, 0.1 mM A␤(1-28) in 50 mM phosphate buffer, in 90% H 2 O, 10% D 2 O, at 25°C. The pH-dependent shifts for ⑀ and ␦ protons were fitted to a modified Hill equation to determine pK a for His 6 , His 13 , and His 14 and are also presented in supplemental Fig. S7.

RESULTS AND DISCUSSION
Copper 2ϩ and Fibril Growth Rates-Using the well established amyloid-binding ThT fluorescence assay, we have investigated fiber formation over a range of A␤ concentrations, with and without the presence of Cu 2ϩ ions. A␤ concentrations of Ͼ10 M, pH 7.4, showed no detectable amyloid fibrils in the presence of 1 mol eq of Cu 2ϩ ions (see supplemental Fig. S1), as reported previously (42,43). However, under more dilute conditions, with A␤ between 5 and 2 M, rapid fiber formation was detected in the presence of Cu 2ϩ ions. Fig. 1 shows that Cu 2ϩ ions significantly increase the rate of A␤ fiber formation at pH 7.4. In Fig. 1a, multiple ThT fluorescence traces are shown, with and without Cu 2ϩ ions present, Fig. 1b shows normalized data from the mean of nine measurements repeated on two separate occasions (individual fluorescence traces are shown in supplemental Fig. S2). Metal-free A␤ preparations typically take more than 70 Ϯ 2 h to reach half-maximal fluorescence (t 50 ), whereas the same A␤ preparations with 0.5 or 1 mol eq of Cu 2ϩ ions cause fibers to form in nearly half the time, 38 Ϯ 2 h (Fig. 1c). A two-tailed unpaired t test confirms that Cu 2ϩ ions significantly increase fiber growth rates with 99.9% confidence. Kinetic parameters taken from the fiber growth curves are given in Table 1 and  supplemental Table S1.
Inspection of the growth curves indicates that both the nucleation and elongation rate are accelerated by Cu 2ϩ ions for A␤ . However the lag time is particularly reduced by Cu 2ϩ ions, from 49 to 16 h, for example (see Table 1). In the case of A␤(1-42), elongation rates and total fiber content generated are significantly enhanced by the presence of Cu 2ϩ ions (supplemental Fig. S3) whereas the lag times are less affected by Cu 2ϩ ions. We have repeated this fiber growth experiment at a number of pH values (8.0, 8.5, and 9.0), and in each case Cu 2ϩ increases the rate of fiber formation (supplemental Fig. S4). Studies with Cu 2ϩ added as a Cu(glycine) 2 chelate produced identical results (supplemental Fig. S5). It is notable that the total ionic strength is unaffected by the Cu 2ϩ addition, which is constant at 160 mM NaCl.
We then characterized the nature of the Cu 2ϩ -promoted amyloids. TEM images indicate the presence of fibers (Fig. 2). Based on TEM images, under these conditions, the morphology of the fibers generated appears quite similar for fibers formed with and without the presence of Cu 2ϩ ions. We note  DECEMBER 31, 2010 • VOLUME 285 • NUMBER 53 that previous studies by other groups using TEM did not reveal fibers in the presence of Cu 2ϩ because the high A␤ and Cu 2ϩ concentrations used caused amorphous precipitation (42)(43)(44)(45)(46). Furthermore, the Cu 2ϩ -generated fibers are capable of seeding fiber formation of fresh, metal-free A␤ as indicated by a reduction in the lag time (see supplemental Fig. S6).

Cu 2؉ Accelerates Fiber Formation of A␤
How does the presence of Cu 2ϩ ions accelerate the rate of fiber formation? At M concentrations of A␤, Cu 2ϩ does not form crossed-linked species (26,27,50), and the Cu 2ϩ coordination geometry is identical in the monomer and fiber (25,28). This rules out copper bridging to form cross-linked A␤ as a possible mechanism of accelerated fiber formation. Cu 2ϩ coordination may trigger the A␤ misfolding that nucleates fiber assembly; however, the conformational changes in A␤ upon Cu 2ϩ binding are small and outside of the fiber core (25,26). It is, however, well documented that intermolecular selfassociation is strongly influenced by the net charge of the protein. As A␤ approaches its isoelectric point, a pI of 5.3, and an overall neutral charge, its solubility decreases (51,52). We investigated the effect of the net charge of A␤ on the rate of fiber formation more quantitatively by varying the pH and monitoring fiber growth. The growth of A␤ fibers over a range of pH values is shown in Fig. 3a. It is clear that as the pH drops from 8.3 to 5.9 the rate at which fibers form significantly increases. Fig. 3b is a plot of lag times versus pH, lag times (t lag ) reduce from 170 (Ϯ8) h at pH 8.3 to 23 (Ϯ4) h at pH 5.9.
We note that pH dependence of the fiber growth rates bears a strong resemblance to the protonation state of the histidine residues and the N-terminal group within A␤. For direct comparison we determined the protonation state of the three His residues within A␤ (His 6 , His 13 , and His 14 ) over a range of pH values using 1 H NMR chemical shift measurements ( Fig. 3b inset and supplemental Fig. S7). The pK a of the His residues at 25°C is 6.7. It appears the protonation state of the three imidazole rings and the N terminus (pK a 7.9), and consequently the net charge of A␤, is crucial to its amyloidogenicity.
In addition to pH, the binding of metal ions will also perturb the net charge of A␤. It is known that Cu 2ϩ (and Zn 2ϩ ) ions bind to the three histidine residues within A␤ (3, 25-28, 36, 49). At pH 7.4 A␤ histidine residues are predominantly (80%) deprotonated and neutrally charged, thus coordination of a divalent Cu 2ϩ (or Zn 2ϩ ) ion to A␤ histidines adds two positive charges. Adding two positive charges to A␤ at pH 7.4 makes the A␤ peptide complex more neutral in overall net charge, and therefore more prone to self-association, with the result that fiber growth times are almost halved.
The stoichiometric effect of Cu 2ϩ on fiber growth was investigated in more detail. All stoichiometries of Cu 2ϩ up to 1 mol eq caused the rate of fibrillization to increase. Interestingly, substoichiometric amounts of Cu 2ϩ between 0.2 and 0.4 mol eq display the greatest increase in fiber growth rates (Fig.  4). This supports the observation that Cu 2ϩ accelerates nucle-  ation (as well as elongation) as this suggests that substoichiometric amounts of Cu 2ϩ can nucleate fiber formation. Further addition of Cu 2ϩ beyond 1 mol eq caused precipitation of A␤ and markedly reduced the amount of fibers generated (supplemental Fig. S1), as previously noted (53,54). The total amount of fibers generated in the presence (or absence) of Cu 2ϩ ions can be quite variable (supplemental Fig. S8a), presumably reflecting competition between fibril formation and amorphous aggregation. Interestingly, Cu 2ϩ ions tend to increase the total amount A␤(1-42) fiber generated (supplemental Fig. S3). As with the effect of Cu 2ϩ ions on A␤ , it is also notable that at lower pH values the maximal intensity of the ThT fluorescence signal is reduced (supplemental Fig.  S8b). As the pH drops closer to the pI of A␤, formation of amorphous aggregates competes with the rapid formation of ordered amyloids (55). This effect is reduced with dilution, but 2-5 M levels of A␤ are required for reliable timely detection. This concentration-dependent process can be likened to the crystallization of proteins, in which overly precipitative conditions for self-association will cause amorphous aggregates rather than ordered crystals to form. A limitation in previous experiments that showed only amorphous aggregates was the high concentration of A␤ and Cu 2ϩ ions used (42,43) (typically 50 M, 100 M, respectively), much higher than that found in vivo.
Interestingly, accelerated fibril formation appears to be quite specific for Cu 2ϩ ions, Zn 2ϩ ions completely inhibit fiber formation even at 3 M A␤(1-40) (Fig. 5). This may be due to the very different coordination geometry (at micromolar concentration) between the two metal ions. Cu 2ϩ ions form an intramolecular complex with A␤ (25)(26)(27)(28) whereas at micromolar levels current data suggest that Zn 2ϩ will form an intermolecular complex, cross-linking between histidine residues on multiple A␤ molecules (34, 36, 49, 56). These cross-linked Zn 2ϩ -A␤ species will inhibit amyloids forming by interfering with the regular cross-beta assembly.
Copper and A␤ Cell Toxicity-We wanted to relate the ability of Cu 2ϩ ions to promote fiber formation of A␤ to the cell toxicity. We have added both monomeric A␤ and fibrillar A␤ to PC12 cells with and without the presence of Cu 2ϩ ions. Cu 2ϩ ions bound to A␤ are more cytotoxic than A␤ in the absence of Cu 2ϩ ions, whereas the same levels of Cu 2ϩ in the absence of A␤ are not toxic to the cells. Fig. 6a shows that A␤(1-42) fibers incubated with PC12 cells are toxic (40% viability), whereas generation of A␤ fibers in the presence of Cu 2ϩ ions (a half-mol eq in Cu 2ϩ ions) makes the A␤(1-42) fibers considerably more toxic (only 4% viability). This experiment was repeated on a number of occasions, and each time the presence of Cu 2ϩ ions consistently enhanced A␤ toxicity to PC12 cells. Control studies show that the same levels of Cu 2ϩ ions (2.5 M) are not toxic to the cells. Supplemental Fig. S9a shows that free Cu 2ϩ ions over a range of concentrations have no detectable toxic effect; Cu 2ϩ ions were added to the cell medium as CuCl 2 . A further control in which Cu 2ϩ was bound to the nonamyloidogenic Cu 2ϩ -binding fragment, A␤(1-16), was studied. Cu 2ϩ ions when bound to A␤ (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16) are not toxic (supplemental Fig. S9b), showing that the toxic effects of Cu 2ϩ are specific to their interaction with A␤(1-42) and A␤ . The use of A␤(1-16) is a particularly good control as Cu 2ϩ binds to the shorter fragment of A␤ with the same affinity and coordination geometry (25-28), but A␤(1-16) lacks the amyloidogenic region and does not form fibrils.
Next, we were interested in the ratios of Cu 2ϩ to A␤ that caused toxicity. As little as 0.01 mol eq of Cu 2ϩ (100 nM) was found to be significantly more toxic than A␤(1-40) fibers in the absence of Cu 2ϩ ions (Fig. 6b). With 0.1 mol eq of Cu 2ϩ , bound A␤ was even more toxic, whereas 0.5 mol eq of Cu 2ϩ ions were also more toxic than A␤  in the absence of Cu 2ϩ ions. However, at the Cu 2ϩ ratio of 1:1 mol eq, the toxic effects of Cu 2ϩ are lost. Indeed at, 5 mol eq of Cu 2ϩ , relative to A␤ , Cu 2ϩ is actually protective (Fig. 6b). Interestingly, these observations can be directly related to the optimum ratio of Cu 2ϩ -A␤ that generates amyloid fibers, as shown in Fig. 4. We have already shown that substoichiometric amounts of Cu 2ϩ (0.2-0.4 mol eq) are more effective at generating amyloid fibers, whereas Cu 2ϩ levels above 1 mol   DECEMBER 31, 2010 • VOLUME 285 • NUMBER 53 eq will actually inhibit fiber formations (supplemental Fig.  S1b). Clearly, the influence of Cu-A␤ ratios on A␤ toxicity relate closely to the ability of Cu 2ϩ to accelerate (or inhibit) fiber formation.

Cu 2؉ Accelerates Fiber Formation of A␤
Both preformed fibrils (Fig. 6a) and A␤ added as a monomer (Fig. 6b) are toxic to cells. The toxic effects of fibrillar A␤ are apparent within a day of addition to the cell medium. However, addition of monomeric A␤  shows no significant cell cytotoxicity within 2 days of incubation; it is only after this time that the toxic effects are apparent. Fiber growth in the cell medium, detected using a ThT assay, suggests that fibers will take a few days to form, supporting the hypothesis that cytotoxicity observed correlates with the rate at which A␤ fibers are generated.
Over a range of concentrations of fibrillar A␤ , Cu 2ϩ ions increase A␤ toxic effects, shown in supplemental Fig.  S10. Counterintuitively, at lower A␤(1-42) concentrations a greater toxic effect is observed. We suggest that a larger amount of toxic species is present because more protofibrils and fewer amorphous aggregates are generated, at the more dilute concentrations of A␤ . Interestingly, A␤(1-42) at 4.5 M has almost no toxic effect; 89 (Ϯ14)% viability is observed. The addition of Cu 2ϩ ions A␤(1-42) becomes markedly toxic with 19 (Ϯ0.4)% cell viability.
There are a few studies already published investigating the relationship between copper and A␤ cell toxicity (19,20,24,42). These studies appear to show conflicting results; in one study copper promoted toxicity (19), and in another it ap-peared to protect against toxicity (42). In these studies the nature of the A␤ preparation was not well defined (i.e. amorphous aggregate, monomer or fiber), and this may be the source of the discrepancies. Here we are able to correlate the kinetics of fibril formation in the presence of Cu 2ϩ ions with the severity of cytotoxicity. The levels of Cu 2ϩ used relative to A␤ can now explain contradictory observations in the literature. This effect is highlighted in Fig. 6b where substoichiometric levels of Cu 2ϩ significantly reduce cell viability, whereas suprastoichiometric levels are actually protective. Clearly, lower substoichiometric levels of Cu 2ϩ are the more physiologically relevant case, suggesting a role for Cu 2ϩ ions in enhancing A␤ toxicity.
There are potentially two reasons for the enhanced toxicity of A␤ in the presence of Cu 2ϩ ions. Diffusible oligomers of A␤ could bind Cu 2ϩ , resulting in a concentration of Cu 2ϩ ions at the neuronal cell surface, where Cu 2ϩ would generate toxic hydrogen peroxide and hydroxyl radicals. Indeed, A␤ oligomers are found clustered at synaptic terminals (57) and cause memory loss due to synapse failure (58). Redox-active Cu 2ϩ ions released at synaptic terminals will cause lipid peroxidation at the cell membrane and so compromise cell integrity (19,21,59), leading to the neuron loss characteristic of AD. The observation that the antioxidant protein catalase and the Cu 2ϩ chelator metallothionein III are protective strongly supports this hypothesis (18 -20, 24). Alternatively, the Cu 2ϩ ions could promote the formation of protofibrillar/fibrillar A␤ species that are toxic to the cells. The rate of production, quantity, or morphology of the Cu 2ϩ -promoted fibers and protofibrillar oligomers may cause the heightened cytotoxicity. We conclude that if it was simply a matter of reactive oxygen species generation by Cu 2ϩ ions bound to A␤ then one might expect the more Cu 2ϩ present, the greater the toxicity; however, in Fig. 6b we observe that small amounts of Cu 2ϩ ions (0.1 M) are more toxic than 50 times as much (5 M) Cu 2ϩ ions. Thus, the ability of Cu 2ϩ to promote fibers (and by inference protofibrillar species) appears to be the significant factor in reactive oxygen species promoted A␤ cell toxicity.

CONCLUSIONS
Cu 2ϩ -A␤ is more toxic to PC12 cells than A␤ on its own; furthermore, cytotoxic effects are related to the ability of Cu 2ϩ ions to promote amyloid fibers and protofibrils. We suggest that Cu 2ϩ ions increase the rate of fiber formation, at pH 7.4, by causing A␤ to approach its isoelectric point. To put our observation in context, the increase in fiber growth rates measured here due to Cu 2ϩ ions is comparable with that observed for (metal-free) A␤(1-40) mutants associated with familial early onset AD (E22K/G/Q), where a halving of the growth times (t 50 ) of fiber formation is also reported (60). Metals have also been proposed as triggers for other misfolding and assembly diseases such as dialysis-related amyloidosis (61), Parkinson disease (62), and prion diseases (63,64), although it remains to be established whether the mechanisms by which metals induce fibrillization are shared. Our observations provide a rationale for the in vivo observations in Drosophilia and mammals which link the AD phenotype with impaired Cu 2ϩ homeostasis (6,7). It is known that Cu 2ϩ levels in the brain increase with age (2); thus, our observations should refocus attention on loss of Cu 2ϩ homeostasis as a possible risk factor in AD. Cu 2ϩ chelators are being investigated in clinical trials as a potential therapy for AD (2,65,66).