β-Amyloid Fibrils in Alzheimer Disease Are Not Inert When Bound to Copper Ions but Can Degrade Hydrogen Peroxide and Generate Reactive Oxygen Species*

Background: Metal-associated β-amyloid (Aβ) aggregates are implicated in the pathogenesis of Alzheimer disease. Results: Copper bound Aβ(1–42) aggregates, including fibrils, degrade hydrogen peroxide, forming hydroxyl radicals and carbonyls. Conclusion: Copper-bound Aβ fibrils can retain redox activity. Significance: Aβ fibrils bound to copper are not inert end points and may be a source of oxidative stress in the Alzheimer brain. According to the “amyloid cascade” hypothesis of Alzheimer disease, the formation of Aβ fibrils and senile plaques in the brain initiates a cascade of events leading to the formation of neurofibrillary tangles, neurodegeneration, and the symptom of dementia. Recently, however, emphasis has shifted away from amyloid fibrils as the predominant toxic form of Aβ toward smaller aggregates, referred to as “soluble oligomers.” These oligomers have become one of the prime suspects for involvement in the early oxidative damage that is evident in this disease. This raises the question whether or not Aβ fibrils are actually “inert tombstones” present at the end of the aggregation process. Here we show that, when Aβ(1–42) aggregates, including fibrils, are bound to Cu(II) ions, they retain their redox activity and are able to degrade hydrogen peroxide (H2O2) with the formation of hydroxyl radicals and the consequent oxidation of the peptide (detected by formation of carbonyl groups). We find that this ability increases as the Cu(II):peptide ratio increases and is accompanied by changes in aggregate morphology, as determined by atomic force microscopy. When aggregates are prepared in the copresence of Cu(II) and Zn(II) ions, the ratio of Cu(II):Zn(II) becomes an important factor in the degeneration of H2O2, the formation of carbonyl groups in the peptide, and in aggregate morphology. We believe, therefore, that Aβ fibrils can destroy H2O2 and generate damaging hydroxyl radicals and, so, are not necessarily inert end points.

ular oxygen to H 2 O 2 and the latter to ⅐ OH and, hence, can generate these two ROS directly. Moreover, the levels of these metals have been reported to be elevated in the amyloid plaque deposits present in the brains of individuals with AD (21,22).
In our own work, we used the technique of electron spin resonance spectroscopy in conjunction with spin trapping to confirm that A␤ does, indeed, self-generate H 2 O 2 , in our case in the presence of only trace levels of metal ions (23,24). We also used the same technique to examine the time profile for H 2 O 2 generation during the process of A␤ aggregation (for details, see Ref. 25) and found three characteristic phases, i.e. a brief time lag before any H 2 O 2 was observed, followed by a rapid growth and then a slow decay in concentration. This early "pulse" of H 2 O 2 generation was found to coincide with the presence of structures resembling "oligomers" or protofibrils, and its concentration profile (i.e. the variation with time) was in complete contrast to the aggregation profile (as monitored by thioflavin T and immunoassay methods), which showed the typical sigmoidal growth curve observed previously with A␤ and other amyloids (26,27). Thus, one of the fundamental molecular mechanisms underlying the pathogenesis of cell damage in AD could be the direct production of ROS by A␤ during the early oligomer phase of its aggregation (23, 28 -30). The behavior that we observed for H 2 O 2 is classical of that expected for a reaction intermediate formed during a series of consecutive processes. During the phase where concentration is increasing rapidly, the rate of H 2 O 2 production exceeds the rate of removal. The latter two rates are roughly in balance as the curve flattens off, whereas, during the final "decay" phase, the rate of removal exceeds the rate of formation. The loss of H 2 O 2 during the final phase of this reaction profile could be due to at least two mechanisms. One of these is the direct bimolecular reaction of the peroxide with the peptide. The other is the destruction of the peroxide by the transfer of an electron from any of the redox-active metal ions that might be bound to the peptide. In the case of Fe(II) ions, this is the well known Fenton reaction.
This raises the question whether the ␤-amyloid fibrils are actually an "inert tombstone" present at the end of the aggregation process, as some researchers believe, or whether they are actively involved in the production and/or removal of ROS. Our published results have suggested that fully formed ␤-amyloid fibrils do not generate H 2 O 2 (at least within the electron spin resonance detection limit) (25). However, as far as we are aware, there are no studies aimed at determining whether fibrils composed of A␤ can degrade H 2 O 2 in vitro.
The aim of the present investigation was to determine whether A␤ fibrils and other aggregates can degrade H 2 O 2 and whether they could be involved in the generation of ⅐ OH with consequent oxidative damage to the peptide. Therefore, we also looked to see whether we could detect the presence of carbonyl groups within the peptide as a marker of this latter damage.
For aggregation and sampling, 10 mM phosphate buffer (PB) (pH 7.4), with and without test concentrations of metal ions (Cu(II), Fe(II), Fe(III), Mn(II), Zn(II)), was added to A␤ to give a 50 M solution of the peptide, vortexed, and then sonicated for 4 ϫ 30 s and incubated at 37°C. After 144 h, samples were taken for thioflavin T (ThT) and atomic force microscopy (AFM) examination. The rest of the sample was spun in a Beckman airfuge for 1 h at 136,000 ϫ g. The supernatant was carefully pipetted off and tested for aggregate content by ThT. The pellets were resuspended and washed in 10 mM PB (pH 7.4) and then spun in the airfuge for a further 1 h. The supernatants were again removed and the pellets resuspended and incubated at 37°C in 1 mM H 2 O 2 in 10 mM PB for up to 96 h. Samples were taken every 24 h for up to 144 h (Amplex red assay) at 0 and 48 h (coumarin assay) and at 0 and 48, 72, and 144 h (for 2,4-dinitrophenylhydrazine (DNPH) immunoassay). Metal ion concentrations were determined by inductively coupled plasma MS (Thermo Elemental X7) after 200ϫ dilution with 0.1 M HNO 3 .
Thioflavin T-15 M ThT was prepared in 50 mM glycine-NaOH buffer (pH 8.5). This was used to prime a Synergy 2 multilabel plate reader (Biotek). A␤ samples were diluted to 25 M in 10 mM PB, and then 3 ϫ 10-l samples were aliquoted into a Nunc 96-well black microtiter plate. The samples were then injected with 50 l of 15 M ThT, mixed, and read 10 times over 2 min with the plate reader at excitation 450 nm and emission 482 nm (32). Average fluorescence was calculated over the 2-min time course, together with standard deviations, and plotted as relative fluorescence units.
Amplex Red-Stock solutions of 10 mM Amplex red in dimethyl sulfoxide and 1 kilo unit/ml HRP in 10 mM PB (pH 7.4) were prepared and aliquoted into working portions. Each aliquot had 10 s of nitrogen gas introduced into them, and then they were frozen using liquid nitrogen and stored at Ϫ80°C. The Amplex red working solution was prepared at 100 M Amplex red and 200 mM/ml HRP in 10 mM PB (pH 7.4), aliquoted in 1-ml portions, and kept at Ϫ80°C in the dark. For fluorescence measurements, samples were diluted 1:400 in 10 mM PB, and then three 15-l samples were pipetted into a Nunc 384-well black plate. 15 l of Amplex red working solution was then added to each well. The plate was shaken and the fluorescence read on the Victor 2 microplate reader at excitation 563 nm and emission 587 nm (33). Means Ϯ S.D. were converted to concentration of H 2 O 2 and plotted.
Atomic Force Microscopy-2 l of 50 M A␤(1-42) aggregated with either no added metal ions or test concentrations of CuCl 2 and/or ZnCl 2 (12.5 or 50 M) in 10 mM PB (Ϯ metal ions) was diluted by 1:10 in MilliQ water and pipetted onto a piece of poly-L-lysine-coated mica. This was allowed to dry and then washed with MilliQ water. This was then imaged using a Digital Instruments multimode scanning probe microscope using tapping mode, taking at least three 10-m images of each sample, followed by 5-, 2-, and 1-m scans. The images were then processed using WSxM scanning probe microscopy software (34). Lateral measurements of A␤ aggregates were adjusted to account for the "tip broadening" effect using the equation where D is the "true" diameter, r is the tip radius, and d is the measured diameter at half vertical height (35). DNPH Immunoassay for Carbonyl Groups-1 mM H 2 O 2 was added to the aggregated, airfuged, and washed A␤(1-42) pellets, and 2-l samples were taken immediately (time zero) and again after 48, 72, or 144 h and added to 6 l of 10 mM DNPH in 6 M guanidine hydrochloride. These samples were incubated at room temperature for 50 min, giving them a gentle vortex every 10 min. 7 l of each sample was then added to 1.393 ml of 10 mM PB (pH 7.4), vortexed, and frozen at Ϫ80°C. Analysis of the samples for carbonyl groups was done via immunoassay using methods outlined by Dalle-Donne et al. (36). Briefly, 200-l samples were plated onto a clear Maxisorb 96-well plate in quadruplicate and incubated overnight at 4°C. Plates were aspirated and then blocked with 250 l of 2% milk powder in 10 mM PBS, 0.5% Tween at 37°C for 1 h with gentle shaking. Plates were then washed four times in 10 mM PBS, and then 200 l of 1 g/ml HRP-labeled, anti-DNPH mAb in block was added to each well. The plates were incubated at 37°C for 1 h with gentle shaking and then washed four times in 10 mM PBS. 200 l of a 1:500 dilution of europium-labeled streptavidin in 0.05% ␥-globulin, 0.5% BSA, 20 M diethylene triamine pentaacetic acid, 10 mM Tris, and 150 mM NaCl (pH 7.4) was added to each well. The plates were shaken for 10 min, covered in foil, and placed on a rocking table for a further 50 min. After washing four times with 10 mM PBS, 200 l of enhancer solution (PerkinElmer Life Sciences) was added to each well. Plates were shaken again for 10 min and then read using a Victor 2 microplate reader with the settings for the time-resolved DELFIA system. Means Ϯ S.D. were plotted in relative fluorescence units.
Coumarin Assay for Hydroxyl Radical Formation-Aggregated, airfuged, and washed A␤(1-42) pellets were resuspended to final concentrations: 1 mM H 2 O 2 , 0.1 mM coumarin carboxylic acid in 10 mM PB. Reaction mixtures were then incubated at 37°C for 48 h. Three 15-l aliquots were plated onto a black 384-well microtiter plate, and fluorescence was read at excitation 390 nm and emission 460 nm (37). Means Ϯ S.D. were plotted in relative fluorescence units.
Data Analysis-All experiments were carried out with a minimum of three independent repeats. Representative data are shown. Statistical analyses performed were completed via oneway analysis of variance followed by Tukey's honest significant difference (HSD) post hoc test to correct for multiple pairwise comparisons. -In our previous research, we found that the levels of self-generated H 2 O 2 (as determined by electron spin resonance spectroscopy or Amplex red assay) increased during the incubation of A␤  in the absence of any added metal ions, reached a maximum, and then slowly declined as incubation/aggregation continued (16,25,38). The self-generation of this H 2 O 2 is believed to proceed via a twoelectron transfer from a suitable transition metal ion to O 2 (16, 19, 23, 25, 29 -30), with its mechanism detailed by others (39,40). In our case, these metal ions are present at trace levels in the buffers or bound to the peptide during synthesis (24). Some of our results for A␤(1-42) incubated at 25 M over a 48-h period and obtained by Amplex red measurements are shown in Fig. 1. When A␤(1-42) is incubated in the presence of an equimolar concentration of Cu(II), we note an acceleration in the generation of H 2 O 2 over the first few hours. Interestingly, though, there is also a much more rapid deterioration in H 2 O 2 levels at this Cu(II):peptide ratio compared with the curve obtained in the absence of added copper, suggesting that, as incubation proceeds, the later aggregates of A␤(1-42) bound to Cu(II) actively degrade H 2 O 2 . We were intrigued by the increased decay rates in these latter experiments, and, in view of these observations, we examined the ability of A␤(1-42) fibrils to degrade H 2 O 2 , particularly when bound to copper ions. In addition to Cu(II), we also determined the influence of Mn(II), Fe(II), Fe(III), and Zn(II) ions on this process and on carbonyl group formation and of Cu(II) and Zn(II) ions on peptide aggregate morphology (as seen by AFM).

A␤(1-42) Aggregation in the Presence of Different Metal Ions and the Effect of the Resulting Aggregates on Hydrogen Peroxide
Decay and Carbonyl Group Formation-We first studied the aggregation of A␤(1-42) over 144 h in the presence or absence of five different metal ions at a metal ion:A␤(1-42) molar ratio of 1:1 ( Fig. 2A) using the ThT method to detect aggregation. We note a decrease in the final level of fluorescence in the presence of some of the metal ions. This decrease could be due to less fibril formation, although the quenching of the fluorescence in the presence of individual metal ions could also be involved. This equimolar metal ion:peptide ratio was selected following a series of preliminary experiments that suggested that a clear difference between the effects of different treatments on H 2 O 2 degradation and carbonyl group formation was readily observed at this ratio. A␤(1-42) aggregation proceeded in the presence of Mn(II), Fe(II), Fe(III), Zn(II), and Cu(II) ions, or in their absence, to generate ThT-positive aggregates. When these aggregates were centrifuged at high speed in a Beckman airfuge and washed, the supernatants were always ThT-negative, indicating that the ThT-positive peptide aggregates had all been retained in the pellets. In subsequent experiments, this pelleted material was resuspended and incubated with 1 mM H 2 O 2 to determine the effects of mature aggregates on the degradation of H 2 O 2 .
The first supernatant and the subsequent wash were also examined by inductively coupled plasma MS to determine the metal ion content in these samples. The results are summarized in the table in Fig. 2B and clearly show that, in the case of Cu(II) and Zn(II), the supernatant and wash contained only very low concentrations of these metal ions, indicating that the great majority of copper and zinc ions are bound to A␤. This observation is not surprising in view of the latest data for the dissociation constant (K d ) for Cu(II) bound to A␤ , which lies in the range 10 pM to 100 nM (41). Our failure to detect free Cu(II) ions is in agreement with data observed by others (employing two unrelated techniques), who also found the absence of any free Cu(II) ions, thus supporting a very high affinity of A␤  for Cu(II) (20). The K d for Zn(II) is believed to be somewhat lower, at 1-20 M (41). Notably, in our experiments, the other metal ions (Mn(II), Fe(II), and Fe(III)) did not show clear evidence for binding to A␤.
We next examined the ability of resuspended pellets of A␤(1-42) aggregates formed in the presence of different metal ions (Mn(II), Fe(II), Fe(III), Zn(II), and Cu(II)) to destroy 1 mM H 2 O 2 , employing Amplex red to monitor H 2 O 2 levels over further incubation periods of up to 72 h. This concentration of H 2 O 2 was selected following a series of trial experiments aimed at establishing the best conditions for our experiments. It is notable (Fig. 2C) that all of the treatments, apart from Cu(II), showed only a small deterioration in H 2 O 2 levels with time, but when A␤(1-42) was bound to Cu(II) ions at a 1:1 incubation ratio, more than 50% of the H 2 O 2 was lost over a 72-h incubation period.
We then examined the ability of A␤(1-42) aggregates formed in the presence or absence of the different metal ions for evidence of carbonyl group formation. Carbonyl formation is well established and used widely as a measure of oxidative damage via reaction with ROS (36,42). In these latter experiments, resuspended pellets of the A␤(1-42) aggregates were incubated at 37°C with H 2 O 2 for a further 48 h. Significant formation of carbonyl groups was only evident when the peptide was incubated with Cu(II) ions (analysis of variance, F(6, 21) ϭ 174.0, p Ͻ 0.001, Tukey's HSD, p Ͻ 0.001) (Fig. 2D).  These results indicate a substantial loss of H 2 O 2 with incubation. Consequently, to investigate possible oxidative damage to A␤(1-42), we next investigated the formation of carbonyl groups in the peptide (via DNPH immunoassay) at various Cu(II):A␤(1-42) molar ratios. As can be seen in Fig. 3B, the formation of carbonyl groups in the peptide was enhanced when Cu(II) ions were present (maximizing with Cu(II):A␤(1-42) at a molar ratio between 1:5 and 1:2).
Because research by others (43) has indicated that the morphology of the A␤ aggregate formed depends on the Cu(II): peptide ratio, we examined the morphology of our aggregates, prepared at 0, 1:4, and 1:1 molar ratios, by AFM. The resulting images are illustrated in Fig. 3C and clearly indicate that the aggregate morphology was affected by the Cu(II):A␤  ratio. In the absence of any added Cu(II) ions, the aggregates formed showed typical amyloid fibril morphology. Two populations of fibrils were observed, one with an approximate height of 3.71 Ϯ 0.78 nm and a width of 9.19 Ϯ 0.12 nm (n ϭ 19) and the other of approximately double that size with a height of 6.88 Ϯ 0.56 and a width of 18.45 Ϯ 0.50 nm (n ϭ 13), the latter being consistent with two smaller fibrils entwined together. As the ratio of Cu(II):A␤(1-42) was increased to 1:4, some small, amorphous, non-fibrillar "granular" aggregates were seen in addition to the fibrillar material. This fibrillar material has similar proportions to the fibrils generated with no copper added, with an approximate height of 3.96 Ϯ 1.05 nm and a width of 7.91 Ϯ 0.18 nm (n ϭ 14). As carbonyl formation maximized at this Cu(II):A␤(1-42) molar ratio, AFM was also performed to check for carbonyl-induced changes to fibril morphology following incubation with H 2 O 2. The images obtained were similar to those presented in Fig. 3C (at the 1:4 ratio) and showed that fibrillar A␤ remained the predominant form present. At a Cu(II):A␤(1-42) ratio of 1:1, no fibrillar aggregates were present, and only small, non-fibrillar, amorphous aggregates were observed. These aggregates have an approximate height of 1.49 Ϯ 0.41 nm (n ϭ 19). Differences in morphology are clear, particularly when the Cu(II):peptide ratio is 1:1.
Finally, in this part of our research, we sought to obtain information on the redox activity of the peptide bound to Cu(II) ions by the investigation of the possible conversion of H 2 O 2 into ⅐ OH (via coumarin fluorescence assay). The results of this part of our investigation are illustrated in Fig. 3D. As can be seen in this Fig. 3D, the level of ⅐ OH detected increased dramatically as the peptide to Cu(II) ion ratio increased, thus indicating the redox activity of the copper-peptide complex. Also illustrated in Fig. 3D are the results of our control experiments, which indicate that neither H 2 O 2 itself nor the buffer (PB) facilitated the formation of any ⅐ OH in the absence of the peptide or Cu(II) ions.
We designed an additional set of experiments to investigate the influence of copper ions on fibrils preformed in their absence. A sample of A␤(1-42) was preaggregated for 144 h in the absence of any deliberately added Cu(II) ions, incubated for a further 24 h in the presence of Cu(II) ions, and washed. Then, the redox activity toward H 2 O 2 was determined. These experiments were undertaken at three different Cu(II):peptide molar ratios (i.e. no added Cu(II) ions, a ratio of 1:4, and a ratio of 1:1), with the results showing a great similarity in the ability to reduce H 2 O 2 levels at each Cu(II):peptide ratio regardless of whether the Cu(II) ions were present during aggregation or added to preformed fibrils (see Fig. 4).
Influence of the Cu(II):Zn(II) Ratio on H 2 O 2 Levels and Aggregate Morphology-As noted above, high levels of both copper and zinc are present in A␤ plaques, and, as such, these and some other metal ions have been implicated by some researchers in the pathogenesis of AD. However, much of this research has concentrated on the influence of "single" metal ions rather than on aggregates formed in the presence of "multiple" ions. Consequently, in this part of our research, we considered the effect of competitive binding between or, potentially, binding of both Cu(II) and Zn(II) to A␤(1-42) and examined how this might affect H 2 O 2 degradation and aggregate morphology.
Our results for the effect of the Cu(II):Zn(II) molar ratio on H 2 O 2 concentration with incubation at 37°C for up to 72 h are shown in Fig. 5A. Experiments at 50 M Cu(II) and Zn(II):Cu(II) ratios of 0:1, 1:4, and 1:1 are shown in the three left columns of Fig. 5A. These results show that the H 2 O 2 concentration not only declined with the length of incubation but also that this decline was less at higher Zn(II) levels. The three center columns of Fig. 5A show our results at 12.5 M Cu(II) ion concentration but with Zn(II):Cu(II) ratios of 0:1, 1:1, and 4:1. This set of experiments shows the same trend with incubation as observed at 50 M Cu(II), but the decline in H 2 O 2 concentration over time was less at a Cu(II) ion concentration of 12.5 M than that observed at 50 M. Finally, the three right columns in Fig. 5A show the results obtained with zero added Cu(II) ions but with Zn(II) concentrations of 0, 12.5, and 50 M. With both zero Cu(II) and Zn(II) ion concentrations, a very small decline in H 2 O 2 levels with incubation was noted, but when Zn(II) ions were present, there was virtually no change in these levels.
Carbonyl group formation (Fig. 5B) revealed some supporting information. Fig. 5B shows that, as the ratio of Zn(II):Cu(II) was increased from 0:1 to 1:1, the level of carbonyl group formation in the aggregated peptide declined steadily. As zinc increased relative to copper, fewer carbonyl groups were detected, consistent with our results given in Fig. 5A, which shows a reduced loss in H 2 O 2 levels as zinc concentration increases. The observation that, at the highest concentration of Zn(II) to Cu(II), H 2 O 2 levels declined but we detected no obvious increase in carbonyl groups formed may be linked to the sensitivity of this assay or morphological changes that occur at these concentrations of Cu(II) and Zn(II). In either case, Cu(II) bound to the aggregates remains at a sufficient level for bound Cu(II) ions to convert H 2 O 2 to ⅐ OH.
Again, we removed some samples to examine their AFM images, and these are illustrated in Fig. 5C. These latter images  reveal some important information. When only Zn(II) is present, small, amorphous, non-fibrillar (fairly spherical) aggregates were seen and have an approximate height of 5.97 Ϯ 1.40 nm and a width of 27.50 Ϯ 2.73 nm (n ϭ 16). However, Fig. 5C shows that, at a Zn(II):Cu(II) ratio of 1:1, both some non-fibrillar and some fibrillar materials were clearly visible. Two populations of fibers were measured, one with an approximate height of 2.01 Ϯ 0.54 nm and a width 8.00 Ϯ 0.58 nm (n ϭ 11) and the other with a height of 4.62 Ϯ 0.86 nm and a width 14.68 Ϯ 0.90 nm (n ϭ 11), clearly formed via the association of two smaller fibers. One such example is indicated in Fig. 5C (arrows).

DISCUSSION
In an earlier publication, we presented the results of our study of the formation and decay of H 2 O 2 during the incubation of both A␤  and the oxidized form of the ABri peptide associated with familial British dementia (25). In both cases, we were able to show that the self-generation of H 2 O 2 occurred as a short "burst" during the very early stages of aggregation when oligomers were the main species present. After this short burst of activity, H 2 O 2 levels reached a maximum before declining as incubation continued. We speculated that that the decline was either due to a direct reaction of H 2 O 2 with the peptide or due to conversion of H 2 O 2 into ⅐ OH via redox activity that then attacked the peptide.
A closer examination of these earlier results indicates that this decline in H 2 O 2 levels continues when fibrillar aggregates are the majority species present. This decline could be due to the conversion of the some of the peroxide into ⅐ OH via the redox activity of any metal ion bound to the peptide. Consequently, in this work, we focused our research on investigating the possibility that, when A␤(1-42) fibrils are bound to a redox-active metal ion such as Cu(II), the metal ion retains its redox activity and, therefore, is capable of degrading H 2 O 2 and releasing the highly reactive ⅐ OH radical. This highly reactive radical, if formed in the vicinity of the peptide, could inflict substantial oxidative damage on the peptide via a range of reactions.
As a preliminary to our main experiments, we tested the ability of preaggregated A␤ , formed by incubating the peptide with various metal ions, to destroy H 2 O 2 . Our results are summarized in Fig. 2C. A very small decrease in H 2 O 2 levels over 72 h of incubation can be detected when the aggregate is made in the presence of Mn(II), Fe(II), or Fe(III) and also in the absence of any added metal ion, but the most dramatic decline by far is noted when A␤ is incubated with Cu(II). This latter observation is strongly supported by the results of our detection of the formation of carbonyl groups within A␤(1-42) when Cu(II) was bound to the peptide (see Fig. 2D). Consequently, we conclude that oxidation is clearly evident when the aggregates are bound to Cu(II), which must, therefore, remain redox-active in the aggregated peptide.
To examine the retention of redox activity in these preformed aggregates, we extended the incubation period to 144 h and examined the influence of the Cu(II):A␤(1-42) molar ratio on H 2 O 2 levels (see Fig. 3A). We noted that there was a small decline in the peroxide level when only sample preparation lev-els of metal ions were present (i.e. in the absence of deliberately added Cu(II)). However, a considerably greater decline was noted when copper ions were incubated with, and bound to, the peptide. The rate of degradation of H 2 O 2 increased as more Cu(II) was included to a maximum at a Cu(II):A␤(1-42) ratio of 1:1 (Fig. 3A). The level of ⅐ OH generated by the A␤  aggregates formed at different molar ratios of Cu(II) supported these observations (Fig. 3D). Increasing Cu(II) generated more ⅐ OH, with a maximum at 1:1, further indicating that loss of H 2 O 2 was due to conversion to ⅐ OH.
We also examined AFM images of the various samples (Fig.  3C) because we are aware that the morphology of the aggregates made in the presence of Cu(II) changes with the Cu(II):A␤(1-42) molar ratio. In agreement with previous results (for example, see Ref. 43), we observed that, toward a 1:1 ratio, the aggregate becomes amorphous and non-fibrillar. At a 1:4 ratio, fibrils are the predominant aggregate observed. When no added copper ions are present, the fibrils tend to be more intertwined, with both single and multiple fibers (with dimensions approximately double those of the single fibers) being seen. These observations compare very favorably with those published previously by others (43). Our main conclusion from this part of research is that the aggregates formed at different Cu(II):A␤(1-42) molar ratios up to 1:1 retain their redox activity irrespective of their morphology and are capable of converting H 2 O 2 into ⅐ OH. However, of particular relevance to AD, the fibrils formed at lower concentrations of Cu(II) also have this redox activity. In further support of the fact that copper-bound A␤ fibrils within senile plaques may also have this capacity, we showed that Cu(II) ions bound to the peptide are redox-active whether they are present during aggregation or are added to preformed fibrils.
We also found evidence for oxidative damage to the peptide via the formation of carbonyl groups (Fig. 3B). The formation of carbonyl groups is well established as an indicator of the attack on proteins by ROS (36,42). This chemical modification can alter protein secondary and tertiary structure, which can be detrimental to protein functionality. Many carbonylated proteins are targeted for digestion by the 20 S proteasome, and they can inhibit proteasome activity (44). There is also some evidence that they are particularly prone to aggregation (45). Protein carbonylation could, therefore, contribute to neuronal degeneration and dysfunction in neurodegenerative disorders through these mechanisms. However, the role of carbonylation of A␤ in the pathogenesis of AD is unclear.
In this study, carbonyl adducts were not evident in A␤ aggregates prior to the addition of H 2 O 2 independently of their metal ion content and, therefore, of their morphology. This suggests that carbonylation is not a major driving force for A␤ amorphous aggregation or fibrillization under these experimental conditions. Upon addition of H 2 O 2 , the formation of carbonyl adducts maximized at a Cu(II):A␤(1-42) molar ratio of between 1:2 and 1:5, when fibrils are the predominant type of aggregate. This may indicate that the morphology of the aggregates is important in determining whether the ⅐ OH formed then oxidize the peptide to form carbonyls, with fibrillar aggregates being more easily oxidized. However, there are two other alternative explanations for this. First, the conformational arrange-ment of the protein in amorphous aggregates could conceal carbonyl groups that are formed to a greater extent than in the fibrillar aggregates and, therefore, inhibit the reaction between DNPH and the anti-DNPH mAb. The second possibility is that genuine fibrillar aggregates formed in the presence of subequimolar concentrations of Cu(II) adhere more tightly to the microtiter plates than the amorphous aggregates formed with equimolar A␤  and Cu(II) and, therefore, give a higher signal for carbonyl groups. Regardless, the oxidative damage detected here, in the form of carbonyl group formation, confirms our observation that redox activity is retained in A␤  aggregates produced in the presence of up to an equimolar concentration of Cu(II), with substantial conversion of H 2 O 2 to ⅐ OH.
It is known that both copper and zinc ion concentrations are high in the brains of AD patients (for example, see Refs. 46, 47 and the references quoted therein), and the study of A␤ in the presence of these ions has received considerable attention. Most of this research has concentrated on the influence of either Cu(II) or Zn(II) on the aggregation and redox activity of A␤. In comparison, little attention has been focused on the aggregates formed when A␤  is coincubated in the presence of both of these ions (48). Therefore, we investigated preformed A␤  aggregates prepared in the presence of both of these ions and at different metal ion ratios (Fig. 5). Fig. 5A shows how H 2 O 2 levels are influenced by Zn(II):Cu(II):A␤(1-42) molar ratios over incubation periods of up to 72 h. At a Cu(II):A␤(1-42) ratio of 1:1 but varying Zn(II) ratios up to 1:1:1, the loss of H 2 O 2 is substantial but declines as Zn(II) increases. The same observation is true at a Cu(II):A␤(1-42) molar ratio of 1:4, but the deterioration in H 2 O 2 is less noticeable than at 1:1. This may due to competition between redoxactive Cu(II) and redox-inactive Zn(II) for binding sites, with those Cu(II) ions still bound to the peptide retaining their redox activity. It is possible, however, that both Cu(II) and Zn(II) could be bound via an alternate binding mode. Cu(II) is believed to have a second, albeit lower-affinity, binding site (49,50). In addition, a number of alternative binding sites are reported for Zn(II) (discussed in Ref. 41), with evidence suggesting only a partial overlap between Cu(II) and Zn(II) binding sites (41,50,51). In this scenario, the binding of Zn(II) may prevent the redox events enabled by Cu(II) binding via disruption of the A␤ aggregate morphology that supports the redox capabilities of the Cu(II)-A␤ complex. Supporting this, in the complete absence of added Cu(II) there is little or no decline in H 2 O 2 levels with incubation as the Zn(II):A␤(1-42) ratio varies. We also found supporting evidence for some Cu(II) binding to A␤(1-42) from our detection of carbonyl groups in the peptide as the Zn(II):A␤(1-42) ratio varied (Fig. 5B). Here, the formation of carbonyl groups decreased as the ratio of Zn(II) ions to Cu(II) ions increased. Again, we believe that this oxidative damage is a consequence of the conversion of H 2 O 2 to ⅐ OH (via redox activity) and the subsequent attack of this latter radical on the peptide. Notably, our results with both Cu(II) and Zn(II) together are compatible with a recent report showing that Zn(II) can "redox-silence" A␤, in this case by suppressing the copper-dependent formation of H 2 O 2 from A␤(1-42) and rescuing cells from the damaging effects of this peptide (52).
Finally, we find that further supporting evidence for competition between Cu(II) and Zn(II) can be seen in the corresponding AFM images (Fig. 5C). When the Zn(II):A␤(1-42) ratio is 1:1 (but in the absence of added Cu(II)), somewhat spherical amorphous aggregates are seen. This type of aggregate has been noted before in the case of A␤(1-40) (27,53,54). However, the images obtained for samples prepared when both metal ions and the peptide are present at a 1:1:1 ratio show the coexistence of fibrillar and amorphous aggregates.
In the Introduction, we outlined a change in opinion regarding the amyloid cascade hypothesis, with a move toward the belief that the predominant toxic forms of A␤ are small, soluble oligomers. Support for this latter view has been mounting over recent years, and fibrillar forms of A␤ are now seen by many researchers as being relatively non-toxic compared with oligomeric forms, or even as a protective response of the brain. However, our new results, presented here, indicate that, in the presence of bound redox-active ions, particularly Cu(II), mature aggregates, including fibrils, retain their redox activity. Consequently, they remain capable of converting H 2 O 2 to the very reactive ⅐ OH radical. This could subsequently result in oxidative damage to the peptide itself and also to surrounding areas of the brain that are immediately adjacent to senile plaque amyloid deposits. Mature fibrils are not, therefore, simply inert tombstones at the end of the aggregation process. This could help to explain the many "reactive" processes that have been found to occur in the brain in the vicinity of senile plaque amyloid deposits, including abnormal neuritic processes of nerve cells and the recruitment and activation of glial cells (55).
We have drawn together some of our main observations and illustrated them in Fig. 6. In the absence of any added Cu(II) (but in the presence of any Cu(II) that is unavoidably present in the materials used in the preaggregation of our samples), we find (Fig. 6, top left) some loss of H 2 O 2 over a 144-h incubation period. Moving counterclockwise, we note that this loss is much more noticeable when the Cu(II):A␤(1-42) ratio is 1:4 and even more obvious when it is 1:1. Changes in aggregate morphology and a decline in H 2 O 2 levels and in carbonyl group formation are also noted as the ratio of copper ions increases. When the Cu(II):Zn(II) ratio is set at 1:1 in the preaggregation of the peptide, the loss of H 2 O 2 is still considerable, but no loss is noted when only Zn(II) (and no Cu(II)) is present (Fig. 6, top  right).
There is further support for our concepts in a recent publication by Sayre et al. (56), who have shown that ␤-amyloid plaques in situ in human brain sections can convert 3,3Ј-diaminobenzidine into its insoluble brown reaction product in the presence of H 2 O 2 but in the complete absence of any peroxidase enzyme. This shows that the amyloid plaques themselves have peroxidase-like activity, which would be compatible with our observation that A␤ fibrils can destroy H 2 O 2 . Interestingly, neurofibrillary tangles also showed a similar property, suggesting that both plaques and tangles are sites of this type of redox activity (56).
Of course, for our hypothesis to be valid, H 2 O 2 would have to be present in the vicinity of senile plaques in the human brain. However, this peroxide is very diffusible within and between cells and is formed in sizeable quantities by a number of biolog-ical systems, including mitochondria and, especially, dysfunctional mitochondria, which appear to be present in AD and other neurodegenerative conditions (for fuller details, see Ref. 57). Moreover, as noted above, some forms of prefibrillar A␤ are also capable of generating H 2 O 2 .
We believe that our new experiments help to emphasize the growing realization that metal ions (especially copper ions) are important in the pathogenesis of AD, supporting recent research like that by Singh et al. (58). It is now widely established that copper ions bind directly to the peptide (59), influence its aggregation, and modify the morphology of the aggregate (43). The redox activity of the copper-bound peptide converts any physiological H 2 O 2 to ⅐ OH (60), thus acting as a source of ROS.
In conclusion, we have shown that the redox activity of the Cu(II)-bound peptide is retained in fibrillar A␤(1-42) aggregates, which are capable of destroying H 2 O 2 . Amyloid fibrils are, therefore, not necessarily "inert bystanders" at the end of the disease process. It is possible that, in the future, new drugs could be developed that specifically target this A␤ redox activity, and these could be beneficial even in the more advanced stages of AD. Metal ion chelators are already in development as a possible AD therapy (61), but drugs that specifically bind to sites in A␤ fibrils that are involved in this redox activity could be a viable alternative option.