Cu(II) Potentiation of Alzheimer Aβ Neurotoxicity

Oxidative stress markers as well as high concentrations of copper are found in the vicinity of Aβ amyloid deposits in Alzheimer's disease. The neurotoxicity of Aβ in cell culture has been linked to H2O2generation by an unknown mechanism. We now report that Cu(II) markedly potentiates the neurotoxicity exhibited by Aβ in cell culture. The potentiation of toxicity is greatest for Aβ1–42 > Aβ1–40 ≫ mouse/rat Aβ1–40, corresponding to their relative capacities to reduce Cu(II) to Cu(I), form H2O2 in cell-free assays and to exhibit amyloid pathology. The copper complex of Aβ1–42 has a highly positive formal reduction potential (≈+500–550 mV versus Ag/AgCl) characteristic of strongly reducing cuproproteins. These findings suggest that certain redox active metal ions may be important in exacerbating and perhaps facilitating Aβ-mediated oxidative damage in Alzheimer's disease.

Oxidative damage in the neocortex coincides with A␤ accumulation both in Alzheimer's disease (AD) 1 (1) and in A␤ amyloid-bearing transgenic mice (2), but the mechanisms of oxidation are unknown. The possibility that A␤ accumulation causes oxidation, perhaps by radical formation (3), has been explored, but the nature of the chemistry involved in generating A␤-associated oxidation products such as lipid peroxides (4) remains to be elaborated. In culture, A␤-induced neurotoxicity is characterized by elevated cellular H 2 O 2 and is combated by antioxidants such as vitamin E and catalase (5). The origin of the toxic H 2 O 2 is unknown.
Recently, we reported that Fe(III) interacts directly with A␤1-42 and A␤1-40 to produce H 2 O 2 and TBARS formation in a cell-free manner in vitro, through reduction of the metal ion (6), suggesting that a source of the H 2 O 2 that mediates toxicity in cell cultures exposed to A␤ is extracellular. Cu(II)and Fe(III) have been found in abnormally high concentrations in amyloid plaques (Ϸ0.4 and Ϸ1 mM, respectively) and AD-affected neuropil (7), and copper-selective chelators have been shown to dissolve A␤ deposits extracted from AD post-mortem brain specimens (8). Therefore, these metal ions may be important cofactors in A␤-associated oxidative damage. Importantly, we have also reported that the generation of both Cu(II) and Fe(III)-mediated TBARS is greatest for A␤1-42 Ͼ A␤1-40 Ͼ Ͼ rat A␤1-40 (6). This rank order is of interest because it mirrors the relative participation of the peptides in amyloid neuropathology, and because the most active one (A␤1-42) is overproduced in familial AD (9). Rats and mice do not develop amyloid (10), even in mice transgenic for familial-AD linked mutant presenilin that overexpress endogenous mouse A␤1-42 (11), probably due to the three amino acid substitutions in their homologue of A␤ (Arg 5 3 Gly, Tyr 10 3 Phe, and His 13 3 Arg) (12).
Although Fe(III) mediates and potentiates A␤1-40 toxicity in cell culture (13), it is not clear whether this is due to metal interaction with the peptide or due to a nonspecific increase in reactive oxygen species (ROS) generation within the cell. Redox active metal ions, such as Cu(II) and Fe(III), play an obligatory role in generating ROS, and in mediating ROS-induced damage (e.g. the Fenton reaction) (14,15). Similarly, the Cu(II) and Fe(III) enhancement of dichlorofluorescein (DCF)-reactive oxygen species generated by A␤25-35 treatment of post-mitochondrial rat cerebrocortex (16) could be due to catalytically enhanced ROS generation within the tissue, rather than due to metal interaction with the peptide.
Cu(II) causes the peptide to aggregate to a greater extent than Fe(III) (A␤1-42 Ͼ A␤1-40 Ͼ rat A␤1-40) (17), a property that may be related to the relative affinities of the metal ions for A␤. We hypothesize that if such redox active metal ions bind to A␤ peptides with high affinity and become more oxidizing, they may potentiate A␤-induced cytotoxicity. Furthermore, if the redox competence of A␤ is responsible for its neurotoxicity, then toxicity should be greatest for A␤1-42 Ͼ A␤1-40 Ͼ rat A␤1-40. A␤1-42 has been reported to be more neurotoxic than A␤1-40 (18), but a comparison of the neurotoxicity of these three peptides, or the effects of Cu(II) upon the potentiation of their respective toxicities in culture, has not yet been reported to our knowledge.
Here we report that, in the presence of Cu(II), A␤ is indeed redox-competent (A␤1-42 Ͼ A␤1-40 Ͼ Ͼ rat A␤1-40), and that a series of electron transfer reactions occur when Cu(II) binds to A␤, including reduction to Cu(I) and consequent O 2 -dependent, cell-free peroxide formation. These changes correlate with a striking potentiation in the neurotoxicities of the respective A␤ species in cell culture, supporting an extracellular origin for the H 2 O 2 that mediates A␤-induced toxicity. These data suggest that formation of an A␤-copper complex may be a pathophysiological interaction, and a new target for therapeutic interdiction in AD.

EXPERIMENTAL PROCEDURES
Reagents-A␤ peptides 1-40 and 1-42 were synthesized by the W. Keck Laboratory, Yale University, New Haven, CT. Confirmatory data were obtained by reproducing experiments with A␤ peptides synthesized and obtained from other sources: Glabe Laboratory, University of California, Irvine, CA; Multhaup Laboratory, University of Heidelberg; U.S. Peptides, Bachem (Torrance, CA); and Sigma. Rat A␤1-40 was synthesized and purified by the Multhaup Laboratory. For the EPR experiments, A␤1-40 was synthesized at the University of Queensland. A␤1-28 and A␤25-35 were purchased from U.S. Peptides, Bachem, and Sigma. A␤40 -1 was purchased from Bachem, and also synthesized by the Multhaup Laboratory (giving corroborating results). A␤ peptide stock solutions were prepared in water treated with Chelex-100 resin (Bio-Rad) and quantified, according to established procedures (17). Cu(II)-Gly stock solutions were used to prevent metal-hydroxy and metal-oxy polymers that form in neutral metal ion solutions and were prepared by mixing National Institute of Standards and Technology (NIST) standard copper with glycine at metal/ligand molar ratio of 1:6. Other reagents are from Sigma, unless otherwise mentioned.
Metal Reduction Assays-Assays were performed using a 96-well microtiter plate (Costar), based upon a modification of established protocols (19). Polypeptides (10 M) or vitamin C (10 M), Cu(II)-glycine and Cu(I) indicator (250 M), either bathocuproine disulfonic acid (BC) or bicinchoninic acid (BCA, 4,4Ј-dicarboxy-2,2Ј-biquinoline), were coincubated in Dulbecco's phosphate-buffered saline (PBS: 1.19 mM CaCl 2 , 0.6 mM MgCl 2 , 2.7 mM KCl, 1.4 mM KH 2 PO 4 , 137 mM NaCl, 7.68 mM Na 2 HPO 4 , pH 7.4), at 37°C. Absorbances were then measured using a plate reader (SPECTRAmax Plus, Molecular Devices). In control samples, both metal ion and indicator were present to determine the background buffer signal. Absorbance of metal ion and peptide present in the absence of indicator were taken to estimate the contribution of light scattering due to turbidity. The net absorbances (A) were obtained by deducting the absorbances from these controls from the absorbances generated by the peptide and metal in the presence of the indicator. Cu in buffer. The possibility that the formal potentials measured were for surface-adsorbed complexes cannot be excluded, and, in fact, the wave shapes (Fig. 3, lines c and d) are suggestive of involvement by adsorbed species. Further investigation is planned to resolve this issue.
Electron Paramagnetic Resonance-A␤1-40 (0.42 mg) was dissolved in 300 l of PBS buffer (20 mM Na 2 HPO 4 , 20 mM NaH 2 PO 4 , 150 mM NaCl, pH 7.4) to which CuCl 2 (as indicated) was later added. Q-, X-, and S-band EPR spectra were recorded on a Bruker ESP300E EPR spectrometer; the magnetic field and microwave frequency were calibrated with a Bruker ER-035M gaussmeter and an EIP 548B microwave frequency counter. Quantitation of the Cu(II)-A␤1-40 EPR resonance was performed by a comparison method (20) using an X-band dual TE 104 rectangular cavity with TEMPO as reference sample. The individual cavities have different modulation amplitudes, which can be calibrated, and different microwave magnetic strengths (B1), which was overcome by measuring standard and reference samples in each cavity. The doubly integrated areas are proportional to the spin concentration and were normalized for the anisotropic probability Ͻg 1 2 Ͼ and instrument settings. The relative areas can then be used to calculate the concentration of Cu(II)-peptide that produces the residual spectrum. The anisotropic spectra of the Cu(II)-AB1-40 complexes were measured at 130 K using a flow-through cryostat. Spin Hamiltonian parameters were extracted from the spectra with the XSophe/ Sophe computer simulation software suite (21). Buffers were demetallated with Chelex 100 resin.
Cell Culture-Rat embryonic day 17 forebrain neuronal cultures were grown at 95% O 2 , 5% CO 2 , 85% humidity for 4 days in serum-free Neurobasal TM medium with B-27 supplement (Life Technologies, Inc.), 20 M L-glutamate, 100 units/ml penicillin, 0.1 g/ml streptomycin, and 2 mM L-glutamine. On the fifth day (treatment day), the medium was replaced with serum-free Neurobasal TM plus L-glutamine without B-27 supplement (vehicle medium). Stock solutions were mixed in vehicle medium to a final concentration of peptide or copper-glycine. Experimental trials were done in triplicate wells. Viable cells were counted manually using a 1-mm 2 grid (10ϫ objective) stained with either calcein-AM (Live/Dead TM assay, Molecular Probes) or trypan blue. Data were analyzed using one-way analysis of variance followed by post-hoc Student-Newman-Keuls method and/or Student's t test. Significance level was set at p Ͻ 0.05.
For the DCF assay, 5 mM of 2Ј,7Ј-dichlorofluorescin diacetate (Molecular Probes) in 100% ethanol was de-acetylated by 0.01 M NaOH for 0.5 h, 200 units/ml horseradish peroxidase was then added, and the DCF solution was neutralized and diluted to 200 M by PBS before use. Then 20 M DCF, 10 M A␤1-42, and 1 M Cu(II)-glycine were coincubated at 37°C for 20 min in PBS. Catalase (1000 units/ml) with or without heat inactivation (100°C for 30 min) was used to validate the signal. The fluorescent readings were recorded by a Packard 96-well fluorocounter (485 nm excitation; 530 nm emission).
Where the O 2 tension of the buffers were manipulated, the buffer vehicle was continuously bubbled for 2 h at 20°C with 100% O 2 to create conditions of increased O 2 tension, or purged with argon (Ar) to create anerobic conditions, prior to the addition of vitamin C or polypeptide.

RESULTS
To establish whether A␤ reduces Cu(II) to Cu(I), we used three independent methods. In the first approach various A␤ peptides, vitamin C (as a positive control), and other control polypeptides were incubated with Cu(II)-glycine chelate. Cu(I) formation was monitored using a chromogenic Cu(I)-trapping agent, BC. A␤1-42 and A␤1-40 were the only tested peptides found to reduce significant amounts of Cu(II) to Cu(I) (gener-ating 6 and 3 M, respectively) during the 1-h incubation period (Fig. 1). Cu(II) was not significantly reduced (Ͻ1 M) by rat/ mouse A␤1-40, reverse sequence human peptide (A␤40 -1), amylin, insulin, A␤1-28, or A␤25-35 (Fig. 1). Since it has been pointed out that BC could also bind to Cu(II), potentially altering its reduction potential and cause artifactual estimates of Cu(I) (23), we corroborated the apparent Cu(II)-reducing properties of A␤1-40/2 by comparing the assay results using BC to those obtained with another Cu(I) detection agent, BCA. The apparent amounts of Cu(I) generated by A␤1-42 and A␤1-40 using either BC or BCA were in excellent agreement (Fig. 1,  inset).
Secondly, EPR spectroscopy was used to measure residual Cu(II) remaining after incubating stoichiometric ratios of CuCl 2 with A␤1-40. This peptide caused a loss of the Cu(II) signal (76%), in relative agreement with the corresponding Cu(I) detected in the bioassays above. The detection of Cu(I) by BC assay, at levels roughly comparable to those estimated from loss of the EPR signal for Cu(II), suggests that the EPR signal was not disappearing due to formation of antiferromagnetically coupled (S ϭ 0) dicopper species. Experimental EPR S-band spectra (Fig. 2) and X-and Q-band spectra (data not shown) establish that copper binds very tightly to these peptides. Fig. 2A shows that an approximately equimolar mixture of A␤1-40 and CuCl 2 produces a single Cu(II)-peptide complex. The multiple resonance signal shown in the EPR spectrum for copper-peptide ( Fig. 2A) is for a single A␤-bound Cu(II) species that is paramagnetic. Spin quantitation employing a dual mode TE 104 X-band rectangular cavity revealed that this EPR signal accounted for 24% of the added Cu(II), consistent with 76% of the Cu(II) being converted to EPR-silent species, very likely Cu(I). There was no evidence of free, uncomplexed Cu(II) remaining after addition of the peptide, since unbound Cu(II) itself gives a different multiple resonance signal. The loss of 76% of the Cu(II) signal upon incubation with A␤1-40 is compatible with peptide-mediated reduction of Cu(II) to diamagnetic Cu(I), which is undetectable.
Computer simulation of the experimental spectrum of A␤1-40 with an axially symmetric spin Hamiltonian and the g and A matrices (g ʈ , 2.295; g Ќ , 2.073; A ʈ , 163.60, A Ќ , 10.0 ϫ 10 Ϫ4 cm Ϫ1 ) yielded the spectrum shown in Fig. 2B. Expansion of the M I ϭ Ϫ1/2 resonance revealed nitrogen ligand hyperfine coupling. Computer simulation of these resonances indicated the presence of at least three nitrogen atoms. The magnitude of the g ʈ and A ʈ values also suggest a tetragonally distorted geometry, which is commonly found in type 2 copper proteins (24), and together with the Blumberg Peisach plots, are consistent with a fourth equatorial ligand binding to copper via an oxygen rather than a sulfur donor atom. Thus, the coordination sphere for the copper-peptide complex is CuN 3 O 1 .
The third line of independent evidence to support A␤-dependent reduction of Cu(II) was the electrochemical behavior of Cu(II), assessed in the presence and absence of A␤ by cyclic voltammetry (Fig. 3). This revealed that A␤1-42 (line c) gives rise to a voltammetric response with a formal reduction potential of approximately ϩ500 -550 mV (versus Ag/AgCl) in phosphate-buffered saline alone. This is an extraordinarily high positive potential, which suggests a Cu(I) oxidation state that is highly stabilized by the peptide. Our buffers, even after careful preparation and filtration through Chelex-100 resin, were found routinely to possess Ϸ0.1 M copper background contamination, as assessed by inductively coupled plasma analysis (8). To determine whether interaction with this trace quantity of metal ion was responsible for the electrochemical response of A␤, we added CuCl 2 (17 M) to the solution. This increased the magnitude of the response from the A␤1-42 solution (line d), consistent with this potential being characteristic of a copper-A␤ complex. Background voltammograms of the PBS vehicle did not produce such peaks (line a), and Cu(II) in PBS showed only a Cu(II/I) peak in the reduction wave at ϷϪ80 mV (line b) along with the return oxidation wave. This reduction process was not observed when Cu(II) was added to A␤1-42, consistent with the complete reaction of Cu(II) with A␤.
Since Cu(I) can in principle reduce O 2 , we tested A␤ peptides in the presence of Cu(II) (1 M) for direct production of H 2 O 2 . We found that peroxide was indeed formed in these solutions,  (Fig. 4A), paralleling the amounts of metal reduction by the same peptides (Fig. 1). Validation of these results was achieved by coincubating A␤ with catalase, which abolished the H 2 O 2 signal in a dose-dependent manner, and also by performing a corroborating assay using dichlorofluorescein (data not shown).
To investigate whether the formation of H 2 O 2 by A␤ was due to the specific reduction of O 2 , we studied the generation of H 2 O 2 by A␤1-42, A␤1-40, and vitamin C under different O 2 tensions in the presence of 1 M Cu(II) (Fig. 4B). The presence of vitamin C was used as a control measure to estimate the maximum amount of H 2 O 2 that could be detected in the buffer vehicle by the non-protein generation of Cu(I). This experiment confirmed that there was a significant dependence of H 2 O 2 production upon the O 2 tension. The presence of either A␤1-42 and A␤1-40 generated significantly more H 2 O 2 (A␤1-42 Ͼ A␤1-40) than vitamin C under any O 2 tension studied, and generated H 2 O 2 under low O 2 tension where vitamin C produced none. Under ambient and argon-purged conditions in this system, the reduction of Cu(II) alone by the positive control, vitamin C, was insufficient to produce detectable H 2 O 2 . Therefore, A␤ facilitated the reduction of O 2 more than would be expected by the interaction of the Cu(I) reduced by A␤ with passively dissolved O 2 . Hence, A␤ acts not only to reduce metal ions, but also to trap molecular O 2 to form H 2 O 2 . These data also suggest that copper cycles between the oxidized and reduced forms when bound to A␤, since the presence of 1 M Cu(II) was sufficient to produce a catalytic amount ( Ϫ formation from A␤, using xanthine (1 mM) with xanthine oxidase (0.015 units/ml) in PBS as a positive control.
To prove that A␤-mediated H 2 O 2 formation is metal-ion dependent, H 2 O 2 production by A␤1-42 in the presence of copperselective chelators was assayed (Fig. 4C) We tested the consequences of Cu(II)-A␤ interaction upon the survival of primary neuronal cultures. We found that the combination of A␤1-42 with Cu(II)-glycine (each at 10 M) significantly potentiated A␤ neurotoxicity (70% cell death in the presence of copper, 40% in its absence) (Fig. 5A). The presence of catalase slightly rescued the toxicity of the peptide alone (30% cell death), but entirely rescued the fraction of A␤1-42 toxicity that was enhanced by Cu(II), indicating that the potentiation of toxicity induced by the presence of Cu(II) was mediated by H 2 O 2 . Catalase could not completely rescue the toxicity of A␤1-42, even when catalase was used at higher concentrations (2000 and 3000 IU, data not shown). Since Cu(II)-glycine alone was not neurotoxic, these results strongly support the possibility that Cu(II) interaction modifies A␤ leading to enhanced H 2 O 2 -mediated neurotoxicity. The inability of catalase to completely rescue the neurotoxicity caused by A␤1-42 suggests that other neurotoxic mechanisms that are not mediated by H 2 O 2 may contribute up to 25% of the lethality observed.
To confirm that Cu(II)-enhanced toxicity of A␤ was mediated by extracellular H 2 O 2 production, we studied the effects of Cu(II)-glycine supplementation upon the toxicity of the other biologically occurring A␤ species in primary neuronal culture, comparing human A␤1-42 and A␤1-40 to rat A␤1-40 (Fig.  5B). In the absence of additional Cu(II), A␤1-42 was observed to be more neurotoxic than both human and rat A␤1-40, whose lethal effects were indistinguishable and marginal at the concentrations tested (10 -25 M). However, whereas additional Cu(II) (10 M) dramatically increased the toxicities of human A␤1-42 and A␤1-40, additional Cu(II) did not enhance the toxicity of rat A␤1-40. The Cu(II)-induced potentiation of A␤ toxicity therefore followed the relationship of A␤1-42 Ͼ A␤1-40 Ͼ Ͼ rat A␤1-40, which parallels both Cu(II) reduction, and the Cu(II)-mediated generation of H 2 O 2 , by the same peptides. Taken together, these data argue that Cu(II) enhances the neurotoxicity of A␤ substantially through the cell-free generation of H 2 O 2 .

DISCUSSION
The experiments described above establish that Cu(II) is reduced by A␤ peptides, that Cu(I) mediates O 2 -dependent cell-free H 2 O 2 generation, and that these properties directly correlate with the Cu(II)-mediated potentiation of A␤ neurotoxicity in cell culture. Previously, we showed that concentrations of Cu(II) at ϭ 1 M induce the aggregation of A␤ (17) and generation of TBARS reactivity (6). The amounts of aggregation (17), TBARS reactivity (6), production of Cu(I) and H 2 O 2 , as well as Cu(II)-enhanced neurotoxicity, are all greatest when generated by A␤1-42 Ͼ A␤1-40 Ͼ Ͼ rat/mouse A␤1-40. This relationship also correlates with the relative participation of these peptides in amyloid pathology (10,25), and in familial AD-associated A␤ processing (9,26).
The absence of free Cu(II), determined by our EPR and electrochemical studies following addition of A␤ (Figs. 2 and 3), indicates that both A␤1-40 and A␤1-42 have a high affinity for Cu(II). In agreement with these findings, we recently estimated the K d for Cu(II) binding to A␤1-42 (Ϸ10 Ϫ15 M) and A␤1-40 (Ϸ10 Ϫ10 M). 2 These observations strongly suggest the possibility of a mediating role for even traces of redox active metal ions in the neurotoxic properties of A␤1-40 and A␤1-42. Addition of Cu(II) and A␤ to cell cultures (Fig. 5, A and B) certainly increased the observed neurotoxicity. Since most of the added Cu(II) would be bound by the peptide within the culture medium, we can conclude that neurotoxicity was mediated via peptide-Cu(II) interactions rather than the effects of free Cu(II) upon the cells.
The capacities of biological A␤ peptides (A␤1-42, A␤1-40, and rat A␤1-40) to reduce Cu(II) and to generate cell-free H 2 O 2 (Figs. 1 and 4) correlated with Cu(II)-enhancement of the peptides' respective toxicities (Fig. 5B). This suggests that the enhanced toxicity caused by Cu(II) supplementation was mediated by cell-free H 2 O 2 production, especially since the increase in toxicity caused by the presence of additional Cu(II) was rescued by catalase (Fig. 5A). However, we also observed that there was a component of neurotoxicity that could not be rescued by catalase (Ϸ25%) and therefore may not have been mediated by cell-free H 2 O 2 production (Fig. 5A). Our data indicate, therefore, that A␤1-40/2 toxicity has a lethal compo-2 C. S. Atwood, R. C. Scarpa, R. E. Tanzi, and A. I. Bush, unpublished observations. nent that is H 2 O 2 -mediated, and a lethal component that is not Recently, we found that Fe(III) is also reduced by A␤ (1-42 Ͼ 1-40 Ͼ rat A␤1-40) (6), but our current data indicate that Cu(I) production under the same conditions, from the same concentrations of metal ions and peptide, occurs to a significantly greater extent than Fe(II) production. The stronger ability of A␤ to reduce Cu(II) to Cu(I) is supported by an unusually positive formal reduction potential for the Cu/A␤1-42 complex, which is similar to that of small blue copper proteins (e.g. laccase type 1-blue, ϩ555 mV, versus Ag/AgCl) and oxidases (24), placing it among the extremes of copper-based biological reducing activities.
The observation that A␤1-28 did not reduce Cu(II) (Fig. 1) indicates that at least part of the hydrophobic carboxyl-terminal domain (residues 29 -40, possibly the methionine at residue 35) is critical for the reduction properties of A␤. Redox active cuproenzymes like superoxide dismutase usually bind the active-site Cu(II) within ␤-sheet and ␤-barrel structures (24). The increased ␤-sheet content caused by the two extra hydrophobic residues (Ile and Ala) on A␤1-42 (30) may therefore increase the redox activity of A␤ by enhancing A␤-copper interaction. A␤-copper reduces O 2 to peroxide under conditions of low O 2 tension (Fig. 4B), such as those expected in the brain parenchyma. Since O 2 is preferentially dissolved in hydrophobic environments (14), the hydrophobic carboxyl terminus of A␤ may create a microenvironment that facilitates electron transfer from Cu(I) to O 2 , thereby promoting more H 2 O 2 production by A␤1-42 than by A␤1-40.
There are three ways in which H 2 O 2 can form. A two-electron transfer to O 2 to generate the peroxide anion (O 2 2Ϫ ), a oneelectron transfer generating the superoxide anion (O 2 . ) that subsequently undergoes dismutation to H 2 O 2 , or two one-electron transfers from each of two coppers to a bridging dioxygen molecule resulting in peroxide. The absence of a superoxide intermediate indicates that, as with the case Fe(III)-mediated H 2 O 2 production (6), Cu(II)-mediated H 2 O 2 production from A␤ must occur by either a double electron transfer from Cu(I) to O 2 (generating Cu(III)), or by two successive one-electron transfers. The simultaneous production of Cu(I) and H 2 O 2 by A␤ raises the possibility of the hydroxyl radical (OH ⅐ ) being formed by Fenton chemistry as reported for superoxide dismutase 1 (31), and we have recently determined that A␤ is carbonylated by the decay of OH ⅐ in the presence of excess H 2 O 2 and Cu(II). 3 A key novel finding of this work is that the cell-free generation of H 2 O 2 by A␤ depends upon only substoichiometric (catalytic) amounts of metal ion (d1:10, copper:A␤), which lends credence to the possibility of these reactions occurring in vivo. The concentration of copper in the neocortex is Ϸ80 M, and that released during synaptic transmission is deduced to be Ϸ15 M (33). While free Cu(II) does not exist in the cytosol (34), extracellular Cu(II) may be more readily exchangeable (24). Cu(II) interaction with A␤ may be exaggerated in AD where copper levels are abnormally elevated in the neuropil (7,36), cerebrospinal fluid (37,38), and especially in amyloid where the concentration of copper has been measured at 30 g/g (Ϸ0.5 mM, dry weight) (7). Extraction of copper may be the basis of our recent discovery that copper-selective chelators, such as bathocuproine, facilitate the solubilization of A␤ from deposits in post-mortem AD brain specimens (8). Therefore, the abnormally high concentrations of copper in AD neuropil and in amyloid may reflect a pathogenic neurochemical milieu that both aggregates A␤ and induces H 2 O 2 production. Since H 2 O 2 freely crosses cell boundaries, it is likely that its dissemination from A␤ deposits will contribute to oxygen radical-mediated damage (39,40), and inappropriate apoptotic signaling in AD brain. Copper-dependent production of H 2 O 2 from A␤ accumulations in the brain should now be examined in the context of potential therapeutic intervention in AD.