Dramatic Aggregation of Alzheimer Aβ by Cu(II) Is Induced by Conditions Representing Physiological Acidosis*

The cortical deposition of Aβ is an event that occurs in Alzheimer’s disease, Down’s syndrome, head injury, and normal aging. Previously, in appraising the effects of different neurochemical factors that impact upon the solubility of Aβ, we observed that Zn2+ was the predominant bioessential metal to induce the aggregation of soluble Aβ at pH 7.4 in vitro and that this reaction is totally reversible with chelation. We now report that unlike other biometals tested at maximal biological concentrations, marked Cu2+-induced aggregation of Aβ1–40 emerged as the solution pH was lowered from 7.4 to 6.8 and that the reaction was completely reversible with either chelation or alkalinization. This interaction was comparable to the pH-dependent effect of Cu2+ on insulin aggregation but was not seen for aprotinin or albumin. Aβ1–40 bound three to four Cu2+ ions when precipitated at pH 7.0. Rapid, pH-sensitive aggregation occurred at low nanomolar concentrations of both Aβ1–40 and Aβ1–42 with submicromolar concentrations of Cu2+. Unlike Aβ1–40, Aβ1–42was precipitated by submicromolar Cu2+ concentrations at pH 7.4. Rat Aβ1–40 and histidine-modified human Aβ1–40 were not aggregated by Zn2+, Cu2+, or Fe3+, indicating that histidine residues are essential for metal-mediated Aβ assembly. These results indicate that H+-induced conformational changes unmask a metal-binding site on Aβ that mediates reversible assembly of the peptide. Since a mildly acidic environment together with increased Zn2+ and Cu2+ are common features of inflammation, we propose that Aβ aggregation by these factors may be a response to local injury. Cu2+, Zn2+, and Fe3+ association with Aβ explains the recently reported enrichment of these metal ions in amyloid plaques in Alzheimer’s disease.

A␤, a low molecular weight (39 -43 amino acids) protein that is a proteolytic product derived from the larger amyloid precursor protein (1)(2)(3), is the major component of neocortical amyloid collections in Alzheimer's disease (AD 1 ; see Refs. 4 and 5). As is the case with other amyloid proteins, A␤ originates as a normally soluble and constitutive protein found in biological fluids and tissue (6 -15). A␤ also aggregates to form diffuse amorphous deposits in AD but also following head injury (16 -18) and in healthy aged individuals (19). Combined with other inflammatory proteins such as proteoglycans (20), amyloid Pcomponent (21), and apolipoprotein E (22,23), A␤ is found in the brains of individuals affected by AD and Down's syndrome (DS) as dense extracellular deposits of twisted ␤-pleated sheet fibrils in the neuropil (senile plaques) and within cerebral blood vessels (amyloid congophilic angiopathy; see Refs. 4 and 5). The deposition of A␤, however, is not confined to the brain parenchyma, having been detected in amyloid diseases of the muscle (24 -26), blood vessels (27,28), and in the kidneys, lungs, skin, subcutaneous tissue, and intestine of AD patients (29,30).
Cerebral A␤ deposition occurs in other aged mammals that have the human A␤ sequence (31) but is not a feature of aged rats (32,33). Although soluble A␤ 1-40 is produced by rat neuronal tissue (34), it contains three amino acid substitutions (Arg 3 Gly, Tyr 3 Phe, and His 3 Arg at positions 5, 10, and 13, respectively (32)) that appear to alter the physicochemical properties of the peptide preventing it from precipitating in the neocortex.
A␤ accumulates in the brain in AD and DS in forms that can be resolubilized in water (35,36) but also in forms that require harsher conditions to extract and exhibit associated SDS-resistant polymerization on polyacrylamide gel electrophoresis (35,37). A␤  is the predominant soluble species in biological fluids, whereas A␤ 1-42 , a minor species of A␤ that is more insoluble in vitro, is the predominant species found in plaques and deposits associated with AD and DS (35)(36)(37)(38)(39)(40)(41).
We have pursued studies of the physicochemical properties of synthetic A␤ peptides in order to appraise the potential of various neurochemical environments to induce A␤ deposition. To this end, we had previously found that A␤ is strikingly precipitated by certain metals in vitro, in particular Zn 2ϩ (42)(43)(44)(45). This is important since zinc and other biometals are concentrated in the brain neocortical parenchyma. A recent study, using micro particle-induced x-ray emission analyses of the cortical and accessory basal nuclei of the amygdala, demonstrated that levels of Zn 2ϩ , Fe 3ϩ , and Cu 2ϩ are significantly elevated within AD neuropil compared with control neuropil and that these metal ions are significantly further concentrated within the core and periphery of plaque deposits (46). We have also found that the extraction of A␤ deposits from brain tissue into aqueous buffers is increased in the presence of chelators of Zn 2ϩ and Cu 2ϩ (47), providing further evidence that these metal ions participate in the deposition of A␤ within amyloid plaques.
In common with the appearance of amorphous and amyloid deposits is the observation that altered neuronal H ϩ homeostasis may accompany Alzheimer's disease and head injury. AD is complicated by cerebral acidosis (pH 6.6; see Ref. 48), which may be related to impaired glucose metabolism (49) or to the inflammatory response seen in AD-affected brain tissue (reviewed in Refs. 50 and 51). The release of metal ions such as Cu ϩ and Fe 2ϩ from metalloproteins is induced by a mildly acidotic environment (52)(53)(54)(55)(56)(57)(58). Therefore, we investigated the effect of several bioessential metal ions on their ability to bind and alter the solubility of A␤ 1-40/42 under mildly acidic conditions (pH Ն 6.6) to determine whether there are any unforeseen interactions that may help explain the propensity for A␤ to precipitate under the mildly acidotic conditions anticipated within the metabolically diseased brain parenchyma. A striking and unexpected interaction between pH and Cu 2ϩ was observed for human A␤.

MATERIALS AND METHODS
Reagents and Preparation-Human A␤ 1-40 peptide was synthesized, purified, and characterized by high pressure liquid chromatography analysis (HPLC), amino acid analysis, and mass spectroscopy by the W. M. Keck Foundation Biotechnology Resource Laboratory (Yale University, New Haven, CT). Rat A␤ 1-40 was obtained from Quality Control Biochemicals, Inc. (Hopkington, MA). The HPLC elution profiles of both peptides in these working batches were identified as a single peak. Amino acid analysis of the synthetic peptides indicated no apparent chemical modifications in the amino acid residues. Synthetic A␤ peptide solutions were dissolved in trifluoroethanol (30% in Milli-Q water (Millipore Corp., Milford, MA)) or 20 mM Hepes (pH 8.5) at a concentration of 0.5-1.0 mg/ml, centrifuged for 20 min at 10,000 ϫ g, and the supernatants (stock A␤) used for subsequent aggregation assays on the day of the experiment. Prior to use all buffers and stock solutions of metal ions were filtered though a 0.22-m filter (Gelman Sciences, Ann Arbor, MI) to remove any particulate matter. All metal ions were the chloride salt, except lead nitrate.
The concentration of stock A␤ peptides, bovine serum albumin (BSA; Pierce), aprotinin (Sigma), and insulin (Sigma) was determined by spectrophotometric absorbance at 214 nm (against calibrated standard curves) or by Micro BCA protein assay (Pierce). The Micro BCA assay was performed by diluting 10 l of stock A␤ (or BSA standard) in distilled, deionized water (140 l) and then adding BCA reagent (150 l) to a 96-well plate and measuring the absorbance at 562 nm. The validity of this assay was previously confirmed by amino acid analysis.
Aggregation Assays-To determine the centrifugation time required to completely sediment aggregated proteins, A␤ 1-40 (2.5 M) in 150 mM NaCl and 20 mM Hepes (pH 7.4) was incubated for 30 min at 37°C with no metal, and under aggregating conditions of Zn 2ϩ (100 M), Cu 2ϩ (100 M), or pH 5.5. Reaction mixtures were centrifuged at 10,000 ϫ g for different times or ultracentrifuged at 100,000 ϫ g for 1 h. Centrifugation for Ͼ10 min at 10,000 ϫ g was sufficient to completely sediment Zn 2ϩ -, Cu 2ϩ -, and pH-induced aggregates of A␤ 1-40 when compared with ultracentrifugation (100,000 ϫ g for 1 h). Therefore, centrifugation for 10 -20 min at 10,000 ϫ g was used in subsequent assays to sediment aggregated particles.
To quantitate the effects of different metals and pH on protein solubility, synthetic A␤ stock, and BSA, aprotinin and insulin stocks were diluted to 2.5 M in 150 mM NaCl in either 20 mM glycine (pH 3.0 -4.5), MES (pH 5.0 -6.2), or Hepes (pH 6.4 -8.8) and then incubated (37°C) with or without metal ions for 30 min. The resultant aggregated particles were sedimented by centrifugation (20 min, 10,000 ϫ g), and soluble protein was measured in the supernatant by the Micro BCA protein assay.
Stoichiometry and Binding Analyses of Cu for A␤-Cu 2ϩ concentrations were determined using the spectrophotometric method of Matsuba and Takahashi (59), adapted to microtiter plate volumes. Sample Cu 2ϩ concentrations were determined from a Cu 2ϩ standard curve (0 -100 M in 20 mM ammonium acetate, 150 mM NaCl (pH 7.4)).
Cu 2ϩ -induced spectral changes in A␤ were monitored by incubating A␤  or A␤ 1-42 (5 M) in 20 mM ammonium acetate buffer, 150 mM NaCl (pH 7.4), with and without various concentrations of Cu 2ϩ for 5 min prior to loading onto a quartz microtiter plate. Absorbance was scanned between 200 and 800 nm on a Spectramax Plus Spectrophotometer (Molecular Devices, Sunnyvale, CA). Incubation of Cu 2ϩ with A␤ induced a change in the absorbance profile of the peptide which was maximal at 208 nm. However, subsequent analyses were performed at 214 nm where the spectral shift was still large but the background signal was lower.
By using A␤ aggregation as an end point for Cu 2ϩ binding, the stoichiometry of Cu 2ϩ binding to A␤ 1-40 was determined by incubating A␤ 1-40 (10 M) in 20 mM ammonium acetate, 150 mM NaCl (pH 7.4) for 30 min at 37°C with Cu 2ϩ or Cu 2ϩ -(glycine) 2 (50 M) (60). Reaction mixtures were centrifuged at 10,000 ϫ g for 20 min, and the ratio of Cu 2ϩ binding to A␤ was estimated by determining the concentrations of soluble A␤ and Cu 2ϩ remaining in the supernatant. It was determined that the presence of A␤ did not affect the assay.
Turbidometric Assays-Turbidity measurements, as an assay for aggregation, were performed in a flat-bottomed 96-well microtiter plate (Corning Costar Corp.), and absorbances (405 nm) were measured using a Spectramax Plus spectrophotometric microplate reader (Molecular Devices, Sunnyvale, CA). Automatic 30-s plate agitation mode was selected for the plate reader to evenly suspend the aggregates in the wells before all readings. A␤ 1-40 stock was brought to 10 M (300 l) in 20 mM Hepes buffer, 150 mM NaCl (pH 6.6, 6.8 or 7.4) Ϯ metal ions prior to incubation (30 min, 37°C) and absorbance measurement.
To investigate the reversibility of Cu 2ϩ -induced A␤ aggregation, 25 M A␤ 1-40 and 25 M Cu 2ϩ were mixed in 67 mM phosphate buffer containing 150 mM NaCl (pH 7.4), to a final volume of 200 l. Turbidity measurements were taken at four 1-min intervals. Subsequently, 20-l aliquots of 10 mM EDTA or 10 mM Cu 2ϩ were added to the wells alternatively, and following a 2-min delay, a further four readings were taken at 1-min intervals. After the final EDTA addition and turbidity reading, the mixtures were incubated for an additional 30 min before taking final readings.
To investigate the reversibility of pH-mediated Cu 2ϩ -induced A␤ 1-40 aggregation, 10 M A␤ 1-40 and 30 M Cu 2ϩ were mixed in 67 mM phosphate buffer containing 150 mM NaCl (pH 7.4), and an initial turbidity measurement was taken. Subsequently, the pH of the solution was successively decreased to 6.6 with 0.1 M HCl and then increased back to 7.4 with 0.1 M NaOH. The pH of the reaction was monitored with a microprobe (Lazar Research Laboratories Inc., Los Angeles, CA) and the turbidity read at 5-min intervals for up to 30 min. This cycle was repeated three times.
Modification of A␤ 1-40 -Blockage of histidine residues was achieved by incubating A␤ 1-40 (1.5 mg/ml dissolved in distilled, deionized H 2 O, following centrifugation (10,000 ϫ g for 20 min) to remove undissolved peptide) with 5 mM diethyl pyrocarbonate (DEPC) in 25 mM phosphate buffer (pH 7.4) for 30 min at 4°C (61). Formation of N-carbethoxyhistidine residues was followed by the characteristic increase in absorbance at 240 nm (62). The modified protein was dialyzed extensively to remove excess DEPC, and its ability to aggregate at pH 7.4 and 6.6 with and without Cu 2ϩ , Fe 3ϩ , and Zn 2ϩ was tested as described above. Subsequently, reversal of the histidine modification was achieved by incubating the modified protein with 0.1 M hydroxylamine hydrochloride in 25 mM phosphate buffer (pH 7.4) for 30 min at 4°C. The restored peptide was dialyzed extensively to remove excess hydroxylamine hydrochloride, and its ability to aggregate at pH 7.4 and 6.6 with and without Cu 2ϩ , Fe 3ϩ and Zn 2ϩ was retested.
Western Blot Analysis of the Aggregation of Nanomolar Concentrations of A␤-Stock A␤ 1-40 or A␤  (1 mg/ml) was diluted to a final concentration of 200 ng in 1 ml (44 nM) of buffer containing 150 mM NaCl, 20 mM Tris (pH 7.4 or 6.6), 0.1 M BSA and CuCl 2 (0, 0.1, 0.5, 1 or 5 M), and the samples were mixed and then incubated (30 min, 37°C). The reaction mixtures were then centrifuged at 12,000 ϫ g for 10 min in a fixed angle rotor centrifuge, and the supernatant was carefully removed leaving a pellet of aggregated A␤ on the side of the tube. Sample buffer (30 l; containing 4% SDS, 5% ␤-mercaptoethanol) was added to the tubes which were vigorously mixed, then heated to 95°C for 5 min, and spun briefly. The SDS-extracted samples and a standard containing 50 ng of A␤ in sample buffer were loaded and analyzed by polyacrylamide gel electrophoresis (Tricine gels, 10 -20%; Novex, San Diego, CA), transferred to polyvinylidene difluoride membranes (Bio-Rad), fixed, blocked, and then probed with the anti-A␤ monoclonal antibody 6E10 (Senetek, Maryland Heights, MI) overnight at 4°C. The blot was then incubated with anti-mouse horseradish peroxidase conjugate (Pierce) for 2 h at 22°C and incubated (5 min) with Supersignal Ultra (Pierce) according to the manufacturer's instructions. The chemiluminescent signal was captured for 10 min at maximum sensitivity using the Fluoro-S Image Analysis System (Bio-Rad), and the electronic images were analyzed using Multi-Analyst Software (Bio-Rad). This chemiluminescent image analysis system is linear over 2 orders of magnitude and has comparable sensitivity to film. Aggregated particles were fixed to the membrane (0.1% glutaraldehyde, 15 min), washed thoroughly with Tris-buffered saline, and then probed with the anti-A␤ monoclonal antibody 6E10 (Senetek, Maryland Heights, MI) overnight at 4°C. Blots were then processed as described above.

RESULTS
To determine the effect of mild acidity on the behavior of A␤ in the presence of different metals, A␤ 1-40 was incubated with different bioessential metal ions at pH 6.6, 6.8, and 7.4 at total metal ion concentrations observed in serum (Fig. 1, A and B). Peptide aggregation was measured by sedimentation (Fig. 1A) and turbidometry (Fig. 1B), which generally gave confirmatory results. Incubation of A␤ in the absence of metal ions induced no detectable aggregation over the pH range tested. Only incubation with Zn 2ϩ or Cu 2ϩ induced Ͼ30% aggregation under these conditions (Fig. 1A). As the [H ϩ ] was increased, all other metal ions induced an observable increase in sedimented A␤ (Fig. 1A), but Cu 2ϩ induced the most striking increase in A␤ aggregation as measured by both sedimentation and turbidometry. Fe 3ϩ induced a marked increase in turbidity (Fig. 1B), but this aggregate was not as readily sedimented as the Cu 2ϩ -or Zn 2ϩ -induced A␤ aggregates (Fig. 1A), probably due to decreased density of the assembly. As expected, Zn 2ϩ induced ϳ50% of the soluble peptide to sediment over the pH range tested, but increasing the [H ϩ ] caused a decrease in Zn 2ϩinduced turbidity (Fig. 1B) which may be due to an alteration in the size or light scattering properties of the Zn 2ϩ -induced aggregates. Incubation of the mixtures for 16 h did not induce any appreciable change in the turbidometry readings compared with the original readings (data not shown), suggesting that the aggregation reactions had reached equilibrium within 30 min.
The effects of higher metal ion concentrations were also compared at equimolar (30 M) concentrations (Fig. 1C). Under these conditions, as the [H ϩ ] increased, Ni 2ϩ induced A␤ aggregation (ϳ85%) which was comparable to the effect of incubation with the same concentration of Cu 2ϩ . Co 2ϩ induced a comparable level of A␤ aggregation to Zn 2ϩ (ϳ70%) and, like Zn 2ϩ , Co 2ϩ -induced aggregation was independent of [H ϩ ]. At 30 M, Fe 3ϩ , Al 3ϩ , and Pb 2ϩ were observed to induce partial (ϳ30%) A␤ aggregation at pH 6.6 but no significant aggregation at pH 7.4.
To quantify the binding affinity of Cu 2ϩ for A␤, spectral analysis was performed. The half-maximal binding of Cu 2ϩ for A␤ 1-40 and A␤ 1-42 based upon shift in absorbance was estimated as 4.0 and 0.3 M, respectively (Fig. 2, A and B). The stoichiometry of Cu 2ϩ binding to A␤ 1-40 in the Cu 2ϩ -induced A␤ aggregate was 3.4 and 3.0 at pH 7.0 and 6.6, respectively.
The stoichiometry of Cu 2ϩ binding to A␤ 1-40 required for precipitation is in agreement with that required for the saturation of the spectral shift ( Fig. 2A). Cu 2ϩ supplied as a Cu 2ϩ -(glycine) 2 complex did not alter the amount of A␤ aggregation or binding stoichiometry compared with that induced by noncomplexed Cu 2ϩ in the same buffer (data not shown).
Next, we tested other proteins to determine whether Cu 2ϩinduced protein aggregation under mildly acidic conditions was specific for A␤ (Fig. 3A). In the absence of metal ions, BSA aggregation was low (Ͻ15%) at both pH 7.4 and 6.6. However, in contrast to A␤ 1-40 , the amount of BSA aggregation increased only marginally in the presence of Cu 2ϩ at both pH 7.4 and 6.6. Aprotinin was significantly aggregated (ϳ30%) by mildly acidic conditions, but unlike A␤ its aggregation at pH 6.6 was not potentiated by the presence of Cu 2ϩ . In contrast, insulin was not aggregated by pH 6.6 or by Cu 2ϩ at pH 7.4. However, there was a marked increase in insulin aggregation at pH 6.6 in the presence of Cu 2ϩ , resembling that seen with A␤ 1-40 .
Since higher concentrations of Zn 2ϩ are required to induce the aggregation of rat A␤ 1-40 compared with human A␤ 1-40 , we examined whether rat A␤ 1-40 also was resistant to Cu 2ϩ -induced aggregation (Fig. 3B). At pH 7.4, there was a similar slight increase in the aggregation of both human A␤ 1-40 and rat A␤ 1-40 as the [Cu 2ϩ ] increased. However, the amount of aggregation of rat A␤ 1-40 as [Cu 2ϩ ] increased at pH 6.6 was less than half that of human A␤ 1-40 under the same conditions. We next proceeded to elaborate the effects of [H ϩ ] upon Cu 2ϩ -mediated precipitation of A␤ over a broader pH range (Fig. 4A). [H ϩ ] alone was sufficient to precipitate A␤ 1-40 (2.5 M) dramatically once the pH was brought below 6.3 (Fig. 4A). At pH 5.0, 80% of the peptide was precipitated, but the peptide was less aggregated by acidic environments below pH 5.0, in agreement with previous reports on the effect of pH on A␤ solubility (reviewed in Ref. 63). Zn 2ϩ (30 M) induced a constant level (ϳ50%) of aggregation between pH 6.2 and 8.5, whereas below pH 6.0, aggregation could be explained predominantly by the effect of [H ϩ ].
In the presence of Cu 2ϩ (30 M), a decrease in pH from 8.8 to 7.4 induced a marked drop in A␤ 1-40 solubility, whereas a slight decrease in pH below pH 7.4 strikingly potentiated the effect of Cu 2ϩ on the aggregation of the peptide (Fig. 4A). Surprisingly, Cu 2ϩ caused Ͼ85% of the available peptide to aggregate by pH 6.8, a pH which plausibly represents a mildly acidotic environment. We hypothesize that conformational changes in A␤ brought about by small increases in [H ϩ ] result in the unmasking of a second metal interaction site that leads to its rapid Cu 2ϩ -dependent aggregation. Below pH 5.0, the ability of both Zn 2ϩ and Cu 2ϩ to aggregate A␤ was diminished, consistent with the observation that Zn 2ϩ binding to A␤ is abolished below pH 6.0 (42), probably due to protonation of histidine residues.
Further elaboration of the relationship between pH and Cu 2ϩ on A␤ 1-40 solubility (Fig. 4B) confirmed that there was potentiation between [H ϩ ] and [Cu 2ϩ ] in producing A␤ aggregation; as the pH fell, less Cu 2ϩ was required to induce the same level of aggregation, suggesting that [H ϩ ] controls Cu 2ϩinduced A␤ 1-40 aggregation. At pH 7.4, Cu 2ϩ -induced A␤ aggregation was far less than that induced by Zn 2ϩ at similar concentrations, consistent with our earlier report (43).
To determine if this reaction occurred at lower concentrations of A␤ and to compare the effects of Cu 2ϩ -induced aggregation of A␤ 1-40 with A␤ 1-42 (which is not stable in solution at micromolar concentrations in aqueous buffers), we incubated both peptides (50 nM) with various Cu 2ϩ concentrations (0 -5 M), sedimented the aggregated peptide, resolubilized the A␤ pellets with SDS, and visualized the A␤ on Western blots. Increased aggregation of A␤ 1-40 was apparent at Cu 2ϩ concentrations as low as 500 nM (Fig. 5A). As previously observed at higher A␤ 1-40 concentrations, a decrease in pH from 7.4 to 6.6 potentiated the effect of Cu 2ϩ on the aggregation of A␤ 1-40 in this system. A␤ 1-42 aggregation followed a similar pattern to that of A␤ 1-40 , but 500 nM Cu 2ϩ appeared to induce more aggregation of A␤ 1-42 than A␤ 1-40 at pH 7.4 (Fig. 5B).
To study this reaction at even lower concentrations of A␤ 1-40, A␤ 1-42 (20 nM), and Cu 2ϩ (Յ1 M), such as those found in cerebrospinal fluid (6,11,13,14), we employed a novel filtration immunodetection system 2 (Fig. 6, A and B). This sensitive technique confirmed that a decrease in pH from 7.4 to 6.6 potentiated the effect of Cu 2ϩ (Ն200 nM) on the aggregation of A␤   (Fig. 6A). An increase in A␤ 1-42 aggregation also was observed with increasing [Cu 2ϩ ] which also was potentiated by pH 6.6 ( Fig. 6B). At pH 7.4, A␤ 1-42 was more sensitive than A␤ 1-40 to aggregation induced by increasing Cu 2ϩ concentrations in this range.
Since transition metals often coordinate to histidine residues, and since rat A␤ 1-40 (His 3 Arg substitution at position 13) exhibits attenuated aggregation by both Cu 2ϩ and Zn 2ϩ , we tested whether modifying histidine residues with DEPC (61) affected metal ion-induced aggregation of human A␤ 1-40 . Treatment of A␤ 1-40 with DEPC resulted in an increase in absorbance at 240 nm compared with the unmodified protein, confirming the formation of N-carbethoxyhistidine residues (data not shown), whereas the formation of O-carbethoxytyrosine, characterized by a decrease in the absorption spectrum at 278 nm, was not observed. N-Carbethoxyhistidine modifica-tion completely abolished A␤ aggregation in the presence of Zn 2ϩ (Fig. 7). Reversal of the modifications with hydroxylamine resulted in almost complete restoration of Zn 2ϩ -induced aggregation of A␤, indicating the absolute requirement for the coordination of Zn 2ϩ to the histidine(s) of A␤ 1-40 in order to induce aggregation. Likewise, the requirement for Fe 3ϩ -histidine coordination for A␤ aggregation was demonstrated by the abolition of Fe 3ϩ -induced A␤ aggregation at pH 6.6 following Ncarbethoxyhistidine modification. As with Zn 2ϩ -mediated A␤ aggregation, reversal of the modifications restored Fe 3ϩ -mediated A␤ aggregation (Fig. 7). A marked decrease in Cu 2ϩinduced aggregation of the modified protein at both pH 7.4 and 6.6 also was observed, and reversal of the modifications restored Cu 2ϩ -induced aggregation, indicating that Cu 2ϩ -histidine coordination also was required for A␤ 1-40 aggregation (Fig. 7).
We have recently reported that Zn 2ϩ -mediated A␤ 1-40 aggregation is totally reversible by chelation, whereas A␤ 1-40 aggregation induced by pH 5.5 is irreversible with alkalinization (45). Therefore, using turbidometry, we examined whether Cu 2ϩ /pH-mediated A␤ 1-40 aggregation was also reversible. We observed that Cu 2ϩ -induced A␤ 1-40 aggregation at pH 7.4 was reversible following EDTA chelation, although for each new aggregation cycle, complete resolubilization of the aggregates required a longer incubation (Fig. 8A).

DISCUSSION
These results indicate that subtle conformational changes mediated by [H ϩ ] alter the interaction of A␤ with transition metal ions, modulating peptide aggregation induced particularly by Cu 2ϩ and Ni 2ϩ over a narrow pH range (6.6 -7.4). This reaction resembles the pH-potentiated, metal ion-induced aggregation that is observed for insulin, in agreement with previous reports (64 -66), but was not a significant feature of the behavior of albumin, aprotinin, or rat A␤   (Fig. 3, A and B). The inability of Cu 2ϩ to precipitate rat A␤ is of interest because the rat does not develop cerebral A␤ deposits with age (32,33), and in the absence of Zn 2ϩ (43) or Cu 2ϩ (Fig. 3B) the solubility of rat A␤ 1-40 is indistinguishable from human A␤  . Interestingly, apolipoprotein E, a protein that is closely associated with A␤ aggregation and implicated in the pathogenesis of AD (67)(68)(69), also binds Cu 2ϩ (70), but its relationship with the reactions described here remains to be determined.
Evidence for an interaction between Cu 2ϩ and A␤ 1-40 has been previously observed. Cu 2ϩ stabilized an apparent A␤ 1-40 dimer on gel chromatography (42), and, in one study, 65 Zn 2ϩ binding to A␤ was displaced by co-incubation with excess Cu 2ϩ (71). Although Clements and co-workers (71) in their study were unable to detect the high affinity binding site for Zn 2ϩ , this may be explained by two major procedural differences in their metal binding analyses compared with ours as follows: first, the absence of a competing transition metal ion in their assay system (Mn 2ϩ in our assay (42)) to abolish low affinity metal binding, and second, the absence of physiological (150 mM) NaCl in their assays which is a factor that we have found profoundly alters the interaction of metal ions with A␤ 1-40 (45). Therefore, we suspect that the displacement of 65 Zn 2ϩ by Cu 2ϩ binding to A␤ described by Clements et al. (71) reflects competition for the less specific, lower affinity Zn 2ϩ -binding site on the peptide. We have previously shown that high affinity Zn 2ϩ binding to A␤ 1-40 , appreciated when Mn 2ϩ is used to create stringent conditions that eliminate lower affinity interactions (42), cannot be displaced by other transition metals. Since Cu 2ϩ has been shown to compete for low affinity metal binding under conditions where high affinity Zn 2ϩ binding would be preserved, we suspect that A␤ 1-40 may be capable of binding Cu 2ϩ and Zn 2ϩ simultaneously, a possibility that may be contingent upon the peptide subunits dimerizing in solution, as we have previously observed (45,72).
Our results support the likelihood of at least two classes of metal-binding sites on A␤ as follows: the Cu 2ϩ /Ni 2ϩ site that is responsible for significant A␤ assembly only under mildly acidic (pH 6.6 -7.0) conditions, and the Zn 2ϩ /Co 2ϩ site that mediates significant A␤ assembly at pH 7.4 (Fig. 1). The finding that A␤ 1-40 associates with Co 2ϩ in a similar manner to Zn 2ϩ is of interest since Co 2ϩ , unlike Zn 2ϩ , is paramagnetic and therefore might be used to substitute for Zn 2ϩ in structural biology studies (73). The interactions of Ni 2ϩ and Co 2ϩ with A␤ were achieved with metal ion concentrations that are supraphysiological for these metal ions, although they are physiologically possible for Cu 2ϩ and Zn 2ϩ . Therefore, we believe that Cu 2ϩ and Zn 2ϩ would be the most relevant interacting metal ions with A␤ in biological systems.
A␤ exhibits spectral alterations in the presence of Cu 2ϩ (Fig.  2), which appear to be due to saturable binding of the metal ion. Like the prion protein (74), we found that A␤ binds multiple copper ions. By using the half-maximal binding of Cu 2ϩ for A␤ as an indicator of affinity, we found that A␤ 1-42 has a higher affinity for Cu 2ϩ than A␤  , which is in agreement with our current findings that A␤ 1-42 is more readily precipitated by Cu 2ϩ at pH 7.4 (Fig. 5). Although the spectral alterations that we observed saturated in response to micromolar Cu 2ϩ concentrations, these changes may correspond to relatively low affinity Cu 2ϩ binding. Since three to four Cu 2ϩ ions bind to each A␤ subunit when the peptide is precipitated by Cu 2ϩ , it is possible that Cu 2ϩ -binding sites of submicromolar affinity exist but could not be appreciated by the current aggregation and spectral assays.
Total Cu 2ϩ has been measured at ϳ15 M in the synaptic cleft and up to 100 M in the normal neocortex (46,(75)(76)(77)(78). However, the concentrations of Cu 2ϩ , Fe 3ϩ , and Zn 2ϩ are each more than doubled (ϳ300, ϳ700, and ϳ790 M, respectively) in the neuropil of the cortical and accessory basal nuclei of the amygdala of AD brains compared with the neuropil of normal age-matched brains (46). These metal ions are even further concentrated within the core and periphery of senile plaque deposits (ϳ400, ϳ950, and ϳ1100 M, respectively) (46). Our present and previous findings would explain a basis for Cu 2ϩ , Fe 3ϩ , and Zn 2ϩ enrichment in cerebral A␤ deposits, which would, in turn become a sink for these metal ions.
The lower levels of A␤ 1-40 turbidity seen at pH 7.4 in the presence of all metals combined (Fig. 1B) compared with that induced by Zn 2ϩ alone may reflect a protective effect of one of the metal ions in the mixture, possibly Mg 2ϩ , which is the only metal ion not to exhibit a precipitating effect even at millimolar concentrations. This protective effect is, however, not evident in incubations performed at lower pH, indicating that Mg 2ϩ may block the pH-insensitive binding site for Zn 2ϩ /Co 2ϩ . This possibility awaits experimental verification.
We have previously described the rapid aggregation of A␤ 1-40 by Zn 2ϩ (42)(43)(44)(45), as well as the abolition of zinc binding to A␤ at pH Ͻ6.0, consistent with histidine-mediated interaction. In the current study, modification of A␤ with DEPC abolished Zn 2ϩ -induced A␤ aggregation (Fig. 7), supporting the possibility that the aggregation of A␤ 1-40 by Zn 2ϩ is dependent upon coordination with histidine. Although Fe 3ϩ induced considerably less aggregation of A␤ than Zn 2ϩ or Cu 2ϩ , even this aggregation reaction appeared to be coordinated by the histidine residues of A␤, suggesting that the association of redox active iron with senile plaques from AD brains recently reported by Smith et al. (79) may be due to the metal ion binding directly to A␤.
Histidine modification did not completely abolish Cu 2ϩ -induced A␤ aggregation, suggesting that a proportion of that form of aggregate is assembled by interaction with non-histidine residues on the peptide. Histidine modification of human A␤ 1-40 reduced the amount of Cu 2ϩ -induced A␤ aggregation to that observed following the incubation of rat A␤ 1-40 with the same Cu 2ϩ concentrations. Taken together, these data indicate that Cu 2ϩ coordination to histidine at position 13 and tyrosine at position 10 may coordinate the aggregation of A␤  . Access to these residues by Cu 2ϩ therefore appears to be restricted at pH 7.4.
The relationship of A␤ 1-40 solubility with pH alone mirrors the conformational changes observed by CD spectra within the N-terminal fragment (residues 1-28) of A␤ (63) as follows: ␣-helical and soluble between pH 1-4 and Ͼ7, but ␤-sheet and aggregated between pH 4 -7. Thus, the altered interaction of Cu 2ϩ with soluble A␤ below pH 7.0 may be due to changes in peptide conformation. A␤ 1-40 aggregation by Cu 2ϩ was observed to be rapidly reversible by alkalinization. Since increasing pH above 7.0 promotes the ␣-helical conformation (63), this conformation may be unfavorable for the Cu 2ϩ coordination that mediates peptide assembly.
Cu 2ϩ -induced A␤ 1-40 aggregation at pH 7.4 also was reversible by chelation; however, less peptide was resolubilized during each cycle perhaps because a denser aggregate was formed during each aggregation cycle which retarded chelator access to the peptide mass. This possibility is supported by the observation that complete resolubilization of the Cu 2ϩ -induced A␤ aggregates by EDTA eventually occurred with additional incubation time. The modulation of A␤ solubility by Cu 2ϩ through a number of aggregation/resolubilization cycles suggests that its interaction does not induce the amyloid ␤-sheet conformation, which is not easily reversible.
Although the pathogenic nature of A␤ in AD is well described (4,5), the function of A␤ remains unclear. Taken together, our results emphasize three physiologically plausible environments that could aggregate A␤ as follows: lowered pH ( Fig. 4 1-7). Inflammation generally induces locally acidic conditions (87,88)  active protein, elements of the complement system, and activated microglial and astroglial cells are observed in AD-affected brains (see Refs. 50 and 51 for reviews). A sustained increase in regional Zn 2ϩ , Cu 2ϩ , and H ϩ concentrations induced by a cortical inflammatory response is predicted by our data to contribute to the deposition of A␤ in AD.
Systemic amyloidoses are often associated with chronic inflammation (97)(98)(99). An example is the amyloidosis caused by serum amyloid A, an acute phase-reactant protein (100). Serum copper levels increase during inflammation, associated with increases in ceruloplasmin, a Cu 2ϩ -transporting protein that transfers Cu 2ϩ to enzymes active in processes of basic metabolism and wound healing such as cytochrome oxidase and lysyl oxidase (101,102), and whose metabolism is abnormal in AD-affected neocortex (103). Whereas the neocortex contains high levels of Cu 2ϩ , the availability of Cu 2ϩ is normally restricted by its binding to metalloproteins such as albumin and ceruloplasmin (104). However, the release of reduced copper from ceruloplasmin is greatly facilitated by acidic environments (52), and periods of mild acidosis may promote an environment of increased "free" Cu 2ϩ and therefore promote its exchange to other regional proteins. Such an exchange of Cu 2ϩ at low pH has been described as mediating the binding of serum amyloid P component, an acute phase reactant that also forms an amyloid, to the cell wall polysaccharide zymosan (105). A similar exchange of Cu 2ϩ within a low pH milieu may be responsible for precipitating a fraction of the brain collections of A␤ in AD.
A further mechanism by which Zn 2ϩ , Cu 2ϩ , and H ϩ levels may be elevated in the cortical interstitium is through abnormal energy metabolism. Intracellular concentrations of Zn 2ϩ and Cu 2ϩ are approximately 1000-and 100-fold higher than extracellular concentrations, respectively (77,78,95). This large gradient between intracellular and extracellular compartments is sustained by highly energy-dependent mechanisms (95, 106). Therefore, alterations in energy metabolism, such as those seen in AD (49), trauma, or vascular compromise, may affect the compartmentalization of these metal ions and possibly promote their pooling in the A␤ compartment. A␤ precipitation by Cu 2ϩ in this case may be potentiated by the acidosis that is associated with abnormal energy metabolism. These mechanisms may contribute to the rapid appearance of A␤ deposits reported to occur following head injury (16,107). The reversible aggregation properties of A␤ at local sites of injury may, in this context, be compatible with a role for the peptide in maintaining regional structural integrity.
If the involvement of Cu 2ϩ in A␤ deposition in AD is confirmed, then the reversibility of this pH-mediated, Cu 2ϩ -induced interaction presents the potential for therapeutic intervention. Thus, cerebral alkalinization or metal ion chelation may be explored as potential therapies for the reversal of A␤ deposition in vivo.