Copper Binding to the Amyloid-β (Aβ) Peptide Associated with Alzheimer's Disease

There is now direct evidence that copper is bound to amyloid-β peptide (Aβ) in senile plaque of Alzheimer's disease. Copper is also linked with the neurotoxicity of Aβ and free radical damage, and Cu2+ chelators represent a possible therapy for Alzheimer's disease. We have therefore used a range of complementary spectroscopies to characterize the coordination of Cu2+ to Aβ in solution. The mode of copper binding is highly pH-dependent. EPR spectroscopy indicates that both coppers have axial, Type II coordination geometry, square-planar or square-pyramidal, with nitrogen and oxygen ligands. Circular dichroism studies indicate that copper chelation causes a structural transition of Aβ. Competition studies with glycine and l-histidine indicate that copper binds to Aβ-(1–28) at pH 7.4 with an affinity of Ka ∼107 m–1. 1H NMR indicates that histidine residues are involved in Cu2+ coordination but that Tyr10 is not. Studies using analogues of Aβ-(1–28) in which each of the histidine residues have been replaced by alanine or in which the N terminus is acetylated suggest that the N terminus and His13 are crucial for Cu2+ binding and that His6 and His14 are also implicated. Evidence for the link between Alzheimer's disease and Cu2+ is growing, and our studies have made a significant contribution to understanding the mode of Cu2+ binding to Aβ in solution.

Alzheimer's disease (AD) 1 is characterized by innumerable deposits of extracellular amyloid plaques. A small peptide, amyloid-␤ peptide (A␤), plays a critical role in the initial build up of these amyloid plaques and is the main constituent of the amyloid deposits (1,2). In addition, genetic alterations underlying familial AD are associated with an increase in the production and/or the deposition of A␤ in the brain (3)(4)(5)(6). Amyloid-␤ peptide can be between 39 and 43 residues in length, of which A␤-  and A␤-  are the most abundant fragments. The N-terminal portion of A␤ is hydrophilic, whereas the C terminus amino acids 29 -42 are rich in hydrophobic residues and represent the transmembrane region in the amyloid precursor pro-tein. The sequence of human A␤-(1-42) is as follows: DAEFR-HDSGYEVHHQKLVFFAEDVGSNKGAIIGLMVGGVVIA.
Soluble A␤-  and A␤-  are found in the cerebrospinal fluid and blood plasma of all humans where A␤-(1-40) has a concentration of 5 nM in cerebrospinal fluid (7,8). It is yet to be established what triggers A␤ to convert from its soluble form to an amyloidogenic form, but it has been shown that physiological levels of Cu 2ϩ and Zn 2ϩ cause marked aggregation of A␤. This process is thought to be the prelude to amyloid formation (9). Levels of these metals are elevated in amyloid plaque deposits: 0.4 mM and 1 mM for Cu 2ϩ and Zn 2ϩ , respectively (10). Cu 2ϩ -induced aggregation of A␤ occurs as the pH is lowered to 6.8. This mildly acidic environment mimics a feature of inflammation found in AD (11). Studies on cerebrospinal fluid indicate that zinc will cause the selective aggregation of endogenous soluble A␤ peptide (12). Metal chelators specific to Cu 2ϩ and Zn 2ϩ will reverse this aggregation process (13,14). In addition, the neurotoxicity of A␤ is linked to metal-induced oxidative damage and is a feature of the pathogenesis of AD (15). A dual role as a pro-and antioxidant in copper redox cycling in a Fenton-type reaction has been proposed for A␤ peptide (16 -22).
There are two commonly expressed objections to the role of copper in AD. The first objection is that AD is not associated with elevated exposure to environmental copper. It is important to clarify this misconception. The total concentration of copper within the brain is potentially more than sufficient to be neurotoxic. As a consequence, the brain has efficient homeostatic mechanisms in place to maintain compartmentalization of metal ions, which when compromised cause neurodegenerative diseases such as Wilson's and Menkes' disease. There is evidence to suggest that homeostatic mechanisms for metal ions are impaired in AD patients, and AD is characterized by altered metal ion-dependent processes and metal ion concentrations in the brain (23). The second objection is that the affinity of A␤ for Cu 2ϩ is too low to bind these metals at their extracellular concentrations. This is also a misconception, since extracellular levels of Cu 2ϩ may reach as high as 15 M (10), whereas we show here that A␤ affinity for Cu 2ϩ is at the submicromolar level and is reported to be much higher in amyloid plaques (24).
A recent study using Raman spectroscopy has provided direct evidence that copper and zinc are bound via the histidine imidazole rings in isolated senile plaque cores (25). Perhaps one of the most significant pieces of evidence to link copper with AD is the observation that normally insoluble amyloid deposits of postmortem brain tissue from AD patients can be solubilized in aqueous media by the presence of metal chelators specific to Cu 2ϩ (14). It has been shown that the use of copper chelators can markedly inhibit amyloid accumulation in AD transgenic mice and are in trials as potential drug therapies for AD (26,27). A recent study has shown that trace amounts of copper in the drinking water of rabbits induce ␤-amyloid plaque accumulation (28,29).
Despite an increasing body of evidence to link Cu 2ϩ with AD, the precise coordination geometry and the residues involved in Cu 2ϩ ligation are yet to be established. Both monomeric and dimeric species have been proposed (30). In addition, there are disagreements as to the affinity and stoichiometry of binding. For example, both attomolar affinities (24) and micromolar affinities (31) for Cu-A␤-(1-42) have been reported. In this study, we use a range of complementary spectroscopies to characterize the binding of Cu 2ϩ to A␤ and the structural changes induced in A␤ upon copper coordination. To facilitate solution spectroscopy methods, we have used the more soluble fragment of A␤-(1-28), which lacks the C-terminal third of the molecule. Residues 29 -42 are highly hydrophobic and are not believed to be associated with direct coordination of the metal ion (30,41). In addition, we have studied a number of analogues of A␤-  in which each of the three histidine residues have been replaced with an alanine.

EXPERIMENTAL PROCEDURES
Peptide Synthesis and Purification-Peptides representing various fragments of the amyloid-␤ peptide were synthesized by employing solid phase Fmoc chemistry and produced by the ABC facility at Imperial College (London, UK). After removal from the resin and deprotection, the samples were purified using reverse phase high pressure liquid chromatography and characterized using mass spectrometry and 1 H NMR.
Titrations-The pH was measured before and after each spectrum was recorded. N-Ethylmorpholine buffer was found not to interfere with Cu 2ϩ binding. Typically, 50 mM N-ethylmorpholine buffer was used for electron paramagnetic resonance (EPR) studies, whereas for 1 H NMR and CD studies, samples were prepared in ultrahigh quality (Ͼ18 megaohms/cm resistivity) water, and the pH was adjusted using small amounts of 0.1 M NaOH or HCl. The peptide concentrations were determined using the extinction coefficient of 1280 M Ϫ1 cm Ϫ1 (due to the single tyrosine residue) (32). Typically, the freeze-dried peptides contained 5-10% moisture by weight. The addition of metal ions or competing ligands to the A␤ peptides was performed using small aliquots from stock aqueous solutions.
Circular Dichroism-CD spectra were recorded on an AVIV Circular Dichroism model 202 spectrometer at 25°C. Typically, a cell with a 0.1-cm path length was used for spectra recorded between 185 and 260 nm with sampling points every 0.5 nm. A 1-cm cell path length was used for data between 240 and 800 nm with a 2-nm sampling interval. A minimum of three scans were recorded, and base-line spectra were subtracted from each spectrum. AVIV software was used to smooth data when necessary. Data was processed using Kaleidagraph spreadsheet/ graph package. Direct CD measurements (, in millidegrees) were converted to molar ellipticity, ⌬⑀ (M Ϫ1 cm Ϫ1 ) using the relationship ⌬⑀ ϭ /33,000 ϫ c ϫ l, where c represents the concentration and l is the path length.
Absorption Spectroscopy (UV-visible)-UV-visible electronic absorption spectra were obtained with a Hitachi U-3010 double beam spectrophotometer, using a 1-cm path length quartz cuvette.
Fluorescence Spectroscopy-Fluorescence spectra were collected using a Hitachi F-2500 fluorescence spectrophotometer. An excitation frequency of 280 nm was used, and data were collected over the range of 290 -400 nm. Samples were placed in a four-sided quartz fluorescence cuvette (Hellma), and data were recorded at room temperature.
EPR-X-band EPR data were recorded using a Bruker ELEXSYS 500 spectrometer, operating at a microwave frequency of ϳ9.3 GHz. All spectra were recorded using a microwave power of 0.63 mW across a sweep width of 2000 G (centered at 3200 G) with a modulation amplitude of 10 G. Samples were frozen in quartz tubes, and experiments were carried out at between 20 and 120 K using a liquid helium cryostat. A minimum of two scans were recorded per spectrum. All EPR spectra shown have been background-subtracted from a water blank with subsequent base-line correction in the XEPR software package using 3rd or 4th order polynomial splines. In order to analyze EPR data using the method described by Peisach and Blumberg (34), it is necessary to convert values from gauss (G) to millikaisers by using the formula A II (millikaisers) ϭ 0.046686 ϫ g ϫ ⌬H, where g ϭ 2.0023 and ⌬H is the A II splitting measured in gauss.
NMR-Proton NMR data were collected using a Varian UNITYplus 600-MHz 1 H frequency spectrometer. Data were processed using the VNMR software package. All spectra were recorded at 25°C in 100% D 2 O solution, typically at a peptide concentration of 1 mM. Proton peak assignments were made by the analysis of two-dimensional total correlation spectroscopy and ROESY spectra of the apo peptides. Total correlation spectroscopy and ROESY spectra were collected with typical mixing times of ϳ75 and ϳ300 ms, respectively.
Design of Peptides-Typically, the N terminus was left as the native amino group, whereas the truncated C terminus was blocked as the ethyl ester at the C terminus. Peptides synthesized of the human sequence are shown in Table I.

RESULTS
pH Dependence of Cu 2ϩ Binding- Fig. 1 shows the EPR spectra of 0.8 mol eq of Cu 2ϩ bound to A␤-(1-28) over a range of pH values between 5 and 10. The EPR spectrum at pH 5 gives a single set of signals typical of type II copper, axial (square-planar or square-pyramidal) coordination geometry. The A II , g II , and g Ќ values are 177 G (ϳ16.5 millikaisers), 2.26, and 2.06, respectively. As the pH is increased, a new set of hyperfine peaks are observed to higher field with A II , g II , and gЌ values of 170 G (ϳ15.9 millikaisers), 2.22, and 2.06, respectively. Commensurate with the appearance of a new set of axial signals, the spectra observed at pH 5 reduces in intensity. At pH 8, the two sets of EPR signals have comparable intensities, and at pH 9, the high field signals dominate. The variation in peak intensity with pH is shown as an inset in Fig. 1. It is clear that at pH 7.4, a mixture of two complexes is observed. Peisach and Blumberg (34) have shown that a combination of A II and g II values can indicate ligand type. The A II and g II values at pH 5 are most typical of three nitrogen and one oxygen ligands (3N1O), although 2N and 2O coordination cannot be ruled out. At pH 10, the A II and g II values are more typical for 4N coordination.
We have carried out complementary studies using CD spectroscopy. Fig. 2 shows visible CD spectra for A␤-(1-28) with one equivalent Cu 2ϩ at various pH values between 4.5 and 10.5. At pH 7.5, a visible absorption band is observed at 600 nm (⑀ 600 nm ϭ ϳ50 M Ϫ1 cm Ϫ1 ), which is typical of a type II Cu 2ϩ d-d transition. At pH 7.5 and below, the associated CD band is extremely weak and is not detected; however, the accompanying CD band at 312 nm assigned as an amide-to-copper charge transfer band is observed at pH 5.5 and above (35). Only as the pH is raised to above pH 8.5 is a negative CD band at 514 nm observed, arising from Cu 2ϩ d-d transitions. The insets in Fig Relatively strong CD bands are often observed for d-d transitions of Cu 2ϩ tetragonal complexes (36,37). However, these complexes involve main-chain amide coordination as well as histidine coordination via the imidazole ring. In these cases, the dominant contribution to optical activity observed is due to the vicinal contributions resulting when the asymmetric ␣-carbon is held in a chelate ring between two chelating donor atoms (e.g. adjacent main-chain amides) (38). At physiological pH and below, the lack of optical activity from the d-d transition of the Cu-A␤ complex suggests that backbone amide coordination is not taking place. It is likely that raising the pH above 8 promotes amide deprotonation and copper coordination by the main chain, resulting in a CD band being observed at 514 nm. In summary, it is clear that A␤ forms a Type II square-planar coordination geometry with Cu 2ϩ , and both EPR and CD measurements indicate that the coordinating ligands are highly pH-dependent, a mixed species is present at physiological pH, and main-chain amide coordination is not present at lower pH values.
Stoichiometry of Cu 2ϩ Binding-EPR spectra of A␤-(1-28) with increasing mol eq of Cu 2ϩ have been collected in order to identify the number of copper ions binding to A␤ and are shown in Fig. 3. With increasing amounts of Cu 2ϩ up to 1 eq, identical line shapes are observed with a commensurate increase in intensity. However, above 1 eq of copper, there is a clear shift to low field for the g Ќ signal; 2.06 at 1 eq of Cu 2ϩ and 2.07 at 2 eq of Cu 2ϩ . The signal increases linearly in intensity up to two equivalents of copper. Comparison of the EPR signal intensities with a Cu(Gly) 2 standard sample confirms that all of the EPR signals for the Cu-A␤ complex are observed and therefore rules out the possibility of EPR silent spin-coupled Cu 2ϩ . The addition of a further mol eq of copper (and up to 5 mol eq) results in no further increase in the intensity of the EPR spectra. Cu 2ϩ ions in water will give a largely EPR silent signal at pH Ͼ7 (39). We have confirmed this observation by obtaining EPR spectra of CuCl 2 in N-ethylmorpholine buffer at pH 7.8; copper in aqueous solution at pH 7.8, 20 K. Double integration of EPR spectra plotted versus mol eq of Cu 2ϩ to peptide (insert) indicates that saturation occurs at 2 mol eq of Cu 2ϩ . EPR spectra for 4.0 and 5.0 mol eq of Cu 2ϩ are not shown for clarity, since they are virtually identical to the 3.0-mol eq spectrum.

FIG. 1. Effect of pH on the EPR spectrum of Cu-A␤-(1-28).
A␤-(1-28) at ϳ50 M in H 2 O with 0.8 mol eq of Cu 2ϩ recorded at 20 K is shown. Changes in the hyperfine splitting features at low field (2600 -3200 G) indicate that the coordinating ligands of the bound Cu 2ϩ center changes with pH, with a mixture of two complexes existing at physiological pH. Peak heights at positions I and II are plotted against pH for both A␤-(1-16) and A␤-(1-28) (inset). The crossover point (50:50 mixture of complexes) is indicated. EPR signals are drastically attenuated relative to Cu 2ϩ bound to A␤. Spin integration of the EPR spectra has been plotted versus copper addition as shown as an inset in Fig. 3. It is clear from the EPR data that A␤-(1-28) binds two Cu 2ϩ ions sequentially. After 2 mol eq of Cu 2ϩ , A␤-(1-28) becomes saturated with copper at physiological pH. Similar studies have been performed using CD, both in the UV and visible regions. Monitoring of the band at 205 nm reveals saturation of A␤-(1-16) and A␤-(1-28) at 2 mol eq of Cu 2ϩ at pH 9. In summary, both EPR-and CD-derived Cu 2ϩ binding curves indicate Cu 2ϩ saturation of binding sites on A␤-(1-28) after 2 mol eq.
Folding of the Backbone in the Presence of Cu 2ϩ -It is believed that amyloid formation in AD and other amyloidogenic diseases such as prion disease are the result of protein or peptide misfolding. The A␤ peptide has a structural transition associated with amyloid formation, with conversion from random coil to an extended ␤ sheet-like conformation (40). CD spectra in the UV region can be used to monitor changes in the main-chain conformation. Fig. 4 shows the CD spectra of A␤-(1-16) and A␤-(1-28) with the addition of increasing amounts of Cu 2ϩ at various pH values. Cu 2ϩ binding to A␤-(1-16) occurs as low at pH 5.9, since the negative signal at 198 nm is reduced in intensity as the pH is increased from pH 5.5 to 5.9 in the presence of Cu 2ϩ . Control experiments with A␤- (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16) in the absence of Cu 2ϩ shows that there are no significant differences in the spectra between pH 5.2 and 9.5 (data not shown). At pH 7.5, the addition of Cu 2ϩ (Fig. 4a) causes a loss of the negative CD band at 198 nm and the appearance of a positive band at 225 nm. The intensity of a second positive contribution at 205 nm is pH-sensitive and is more apparent at higher pH values. (For comparison, see Fig. 4, a and b, which show copper titrations at pH 7.5 and 9.5, respectively.) Similar changes in the CD spectra of A␤-(1-28) with Cu 2ϩ addition are observed as shown in Fig. 4c. A positive band at 225 nm appears and is accompanied by a loss of negative band at 198 nm. The positive contribution at 205 nm with copper is observed but is swamped by the more intense random coil CD band at 200 nm. Aposubtracted difference spectra, which illustrate the effect of the Cu 2ϩ addition to both A␤-(1-16) and A␤-(1-28) are shown as insets in Fig. 4, a-c. The difference spectra emphasize the similarity in the changes in the secondary structure with copper addition between A␤-(1-16) and A␤-(1-28). The additional 12 residues to the C terminus have no effect on the copperinduced conformational transition. The changes in the spectra with the Cu 2ϩ addition are complete by 2 mol eq of Cu 2ϩ , which agrees with the stoichiometries determined by EPR (Fig. 3). Isodichroic points are observed at 217 and 195 nm, and these are maintained between 0 and 1.0 eq of Cu 2ϩ . If a dimeric species were formed, the isodichroic point would be expected to shift at 0.5 eq of Cu 2ϩ .
The chirality at 217 nm is largely invariant with the addition of copper between pH values of 5.5 and 9.5. An increased negative CD signal at 217 nm is often attributed to ␤-sheet or extended conformation. The copper-induced structuring of A␤ can therefore not be directly attributed to an increase in ␤-sheet or extended structure. The appearance of a positive CD contribution at 205 nm is not characteristic of any defined secondary structure but does indicate increased ordering of the main chain.
Affinity of Copper Binding-Key to the physiological significance of Cu 2ϩ binding to A␤ is its affinity. With this in mind, we have used the competitive effects of glycine and L-histidine to measure Cu 2ϩ affinity for A␤ by fluorescence spectroscopy. Fig. 5a shows that the addition of Cu 2ϩ to A␤-(1-28) causes marked quenching of the tyrosine fluorescence signal at 307 nm. As glycine is added, it competes with A␤ for the Cu 2ϩ , and the tyrosine fluorescence signal reappears, as shown in Fig. 5b. Cu 2ϩ coordinates to glycine via the amino and carboxylate groups with an apparent (pH-adjusted) K a ϭ 1.8 *10 6 M Ϫ1 , and two glycine residues will bind to a single Cu 2ϩ ion (33). It takes more than 100 mol eq of glycine to cause the tyrosine fluorescence signal to completely return to its maximal strength. Half of the maximal quenching is achieved at ϳ18 Ϯ 2 eq of glycine. Thus, the affinity of Cu 2ϩ for A␤-(1-28) is at least an order of magnitude higher than that of glycine, putting the dissociation constant K d in the submicromolar range (K d Ͻ Ͻ 0.5 M). Similar experiments have been carried out using L-histidine as the competing ligand. In this case, tyrosine fluorescence returns to its maximal value with only 2.5 mol eq of L-histidine, as shown in Fig. 5c. Two molecules of histidine will bind a single Cu 2ϩ ion using the amino and imidazole nitrogens as ligands with an apparent K d at pH 7.8 of 1.5 nM. This indicates that copper will bind to A␤-(1-28) with a lower affinity than L-histidine. This puts the affinity of Cu 2ϩ for A␤ greater than 1.8 ϫ 10 6 M Ϫ1 but less than 6.7 ϫ 10 8 M Ϫ1 or a K d of Ͻ Ͻ500 nM but Ͼ1.5 nM (i.e. 10 -100 nM).
We have obtained similar Cu 2ϩ affinities for A␤-(1-28) using CD spectroscopy. CD was used to directly measure the copper-A␤-associated absorption band at ϳ314 nm. Using the competitive effects of glycine, copper absorption bands become CDsilent when bound to nonchiral glycine (37). We find using the CD band at ϳ314 nm that similar amounts of glycine are required to remove Cu 2ϩ from A␤-(1-28) as is indicated by the fluorescence quenching experiments.
Very high affinities (10 15 M Ϫ1 ) have previously been suggested for Cu-(A␤) 2 (24) (i.e. at 0.5 eq of Cu 2ϩ ). We note that glycine or L-histidine would have little impact on the binding of copper to such a high affinity site. This could result in a false plateau for the fluorescence quenching data shown in Fig. 5. To rule out this possibility, we have added substoichiometric levels of Cu 2ϩ to A␤ (0.3 mol eq of Cu 2ϩ ). If there is indeed a very high affinity site for copper associated with A␤- (1-28), then the addition of L-histidine would have little effect on the fluorescence signal. However, the addition of just 1 eq L-histidine to A␤-(1-28) with 0.3 mol eq of Cu 2ϩ present caused the fluorescence signal to return to its maximal value. We can also rule out apo-A␤ being inadvertently loaded with Cu 2ϩ during peptide synthesis and sample preparation, since no Cu 2ϩ -associated signals are observed for apo-A␤-(1-28) in EPR or CD spectra.
In summary, using both direct measurements of Cu-A␤-(1-28) from CD absorption bands and indirect fluorescence quenching methods, we have shown that the first molar equivalent of Cu 2ϩ ions bind to A␤ with a dissociation constant in the submicromolar level, 10 -100 nM. The possibility of a high affinity copper site for A␤-(1-28) has been ruled out.
Cu 2ϩ Coordination Ligands-Fragments of A␤ have been used to determine which residues are involved in binding to Cu 2ϩ . The three histidine residues within A␤-(1-42) are thought to be the most likely candidates for Cu 2ϩ coordination under physiological conditions. A␤-(1-28), A␤-(1-16), A␤-(1-11) and A␤-(Ac10 -16) (which is N-terminally blocked) were studied. CD spectra in the visible region and the EPR spectra of Cu-A␤-(1-16) and Cu-A␤-(1-28) are almost indistinguishable. A comparison was made over a range of pH values between 4.5 and 9.5 at 1 mol eq of Cu 2ϩ . In the presence of 3 mol eq of Cu 2ϩ , the visible CD spectra of A␤-(1-16) and A␤-(1-28) are also very similar at all pH values. From Fig. 6, it is clear that residues 17-28 have little influence on the binding of Cu 2ϩ to A␤ either at 1 or 3 mol eq of Cu 2ϩ over a range of pH values. The differences in the spectra of A␤-(1-16) and A␤-(1-28) around 270 nm are due to the presence of two Phe aromatic CD signals in the longer fragment irrespective of the presence of Cu 2ϩ .
In contrast, the CD spectra of A␤-(Ac10 -16) and A␤-(1-11) are very different from A␤-(1-16) and A␤-(1-28), as shown in Fig. 6. In particular, at pH 7.5, A␤-(1-16) and A␤-(1-28) show almost no CD bands associated with d-d transitions above 500 nm, whereas A␤-(1-11) has a negative CD band at 590 nm and A␤-(Ac10 -16) has a strong positive CD band at 560 nm. The CD spectra of A␤-(Ac10 -16) suggests that the N terminus and/or His 6 are involved in copper coordination. In addition, A␤-(1-11) CD spectra indicate that His 13 and/or His 14 are key residues in coordinating the Cu 2ϩ ion. A difference in the EPR spectra between Cu-A␤-(Ac10 -16) and copper complexes of the larger fragments are observed but are less pronounced, since A␤-(Ac10 -16) still gives rise to an axial Type II Cu 2ϩ spectrum (data not shown). The pH dependence of binding is similar for A␤-(1-16) and A␤-(1-28), although EPR data (see the inset of Fig. 1) indicates that the transition between the pH 6 and 9 coordination mode occurs at a higher pH for A␤-(1-16) with a midpoint of 8.7 rather than 8.0.
A␤-(1-28) Analogues A␤(H6A), A␤(H13A), A␤(H14A), and A␤-(Ac1-28)-To identify the residues directly involved in the coordination of copper, three analogues have been synthesized in which each of the histidine residues has been replaced with an alanine residue. An additional peptide has also been synthesized in which the N terminus is blocked by acetylation. It is believed that nitrogens from these four loci are the most likely candidates for copper coordination. The EPR spectra of all four analogues studied indicate axial Cu 2ϩ coordination containing nitrogen and oxygen ligands. At pH 6, differences in the spectra between all of the analogues and A␤-(1-28) are slight; at pH 9, the Cu-EPR spectra of the analogues are largely unchanged. In contrast, as shown in Fig. 1, there is a strong pH dependence to the Cu-EPR spectra of wild-type A␤-(1-28). The EPR data suggest that the complex formed at the higher pH value requires the presence of all of the His residues and the N-terminal amino group. However, we have used a number of further spectroscopic techniques described below to support this assertion. For example, when coordination geometry and ligands are similar, visible CD can potentially be much more sensitive to the coordination sphere around the Cu 2ϩ ion compared with EPR.
Visible CD spectra are very sensitive to the relative position of coordinating ligands with changes in the intensity and sign of CD bands (38). The visible CD spectra of the wild-type peptide and CD spectra for the four A␤-(1-28) mutants are shown in Fig. 7. Spectra have been obtained at pH 5.5, 7.5, and 9.5 for all four analogues as well as wild-type A␤- . In addition, spectra have been obtained with both 1 and 3 mol eq of Cu 2ϩ present. CD bands associated with the Cu 2ϩ ion are observed at key positions of 514 nm (assigned to copper d-d transitions), 360 nm (assigned to N im Ͼ Cu 2ϩ charge transfer), 312 nm (assigned to nitrogen amide to Cu 2ϩ charge transfer), and 280 nm (assigned to N im 2 Ͼ Cu 2ϩ charge transfer) (35). Taking each analogue in turn with 1 eq of Cu(II) at pH 9.5, the N-terminally blocked A␤-(Ac1-28) gives a very different trace to wild-type A␤-(1-28), implying that the N terminus is essential for Cu(II) binding. In particular, only A␤-(Ac1-28) gives a CD band at 360 nm, and the band at 514 nm increases by ϳ35% relative to the wild-type spectrum. Furthermore, only A␤-(Ac1-28) gives a positive chirality above 580 nm. The spectra of A␤(H13A) are also perturbed relative to the wild-type spectra. In particular there is a ϳ40% drop in signal at 514 nm but with an increase in the positive charge transfer band below 280 nm. The copper-loaded spectrum of A␤(H14A) has some differences relative to the wild type, with small reduction in intensity (ϳ17%) of the band at 514 nm. It is not clear if the small drop in intensity for the band at 514 nm is significant or due to reduced solubility or small variation in pH. Similarly, the differences between A␤(H6A) and the wild-type spectrum are slight, with a very small reduction (ϳ4%) in the band at 514 nm. Taking these observations together, it appears that at 1 eq of Cu 2ϩ , His 13 and the N terminus certainly coordinate the copper ion. Cu 2ϩ coordination to His 14 and His 6 is less apparent from the visible CD data but cannot be ruled out. Indeed, it is apparent from the CD data in the UV region as shown in Fig.  8 and described below that His 14 and His 6 are involved in coordination.
Copper titrations on each of the analogues were also studied using CD in the UV region and compared with the wild-type spectra. Fig. 8 shows the apo-subtracted difference CD spectra of each of the four analogues and the wild-type spectra after 1 eq of Cu 2ϩ was added. All four analogues look significantly different from wild-type A␤-(1-28) and A␤- (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16). For example, none of the analogues show positive chirality at 225 nm. In addition, the other positive CD band at 205 nm is less intense for the analogues relative to the wild-type and is shifted to 200 nm. This finding therefore implicates not only the N-terminal amino group in the coordination of the first mol eq of Cu 2ϩ but also all three histidine side chains.
In a further attempt to confirm the ligands directly involved in copper coordination, the affinity for each of the analogues was measured using glycine competition and fluorescence (as , and wild-type A␤ (1-28) with the addition of 1 mol eq of Cu 2ϩ , pH ϳ7.6. CD spectra are apo-subtracted difference spectra. described previously). Fluorescence quenching in the presence of competing glycine indicates that A␤-(1-16) and A␤-(1-28) have similar affinity for copper, with glycine causing a return to half-maximal fluorescence signal at ϳ18 Ϯ 2 mol eq of glycine. N-blocked A␤-(1-28) shows a sharp reduction in affinity for copper. This supports the CD studies that indicate the N terminus is a key ligand for copper binding. A␤-(Ac1-28) gives a half-maximal fluorescence signal after the addition of between only 5 and 8 eq of glycine. Similarly, A␤-(Ac10 -16) gives a half-maximal fluorescence signal at 3 eq of glycine, implying that A␤-(Ac10 -16) has an affinity for copper only comparable with the coordination of glycine. The analogue with the His 6 removed also shows a reduction in affinity for Cu 2ϩ with only ϳ11 mol eq of glycine required to produce a half-maximal fluorescence. A␤-(1-28) analogues with His 13 or His 14 removed show a smaller reduction in affinity relative to the wild-type A␤-(1-28), with a half-maximal return of fluorescence with ϳ14 -16 mol eq of glycine in each case. In agreement with the UV-CD data shown in Fig. 8, the affinity of Cu 2ϩ for A␤-(1-28) is reduced by blocking of the N terminus or the removal of any of the three histidine residues.
1 H NMR-1 H NMR has been used to isolate coordinating ligands still further. The apo spectrum of A␤-(1-28) gives line widths indicative of a monomeric peptide. Paramagnetic Cu 2ϩ will broaden 1 H NMR signals when in close proximity to (Ͻ7 Å) or directly coordinated to the metal ion. Fig. 9 clearly shows profound broadening of the H⑀1 and H␦2 resonances for all three histidine residues of A␤-(1-28) at pH 7.8 relative to the other aromatic side chains. In particular, Tyr 10 , which has been suggested as a potential ligand to Cu 2ϩ (30,41), is relatively unaffected by Cu 2ϩ addition and for this reason is unlikely to be involved in direct coordination to the Cu 2ϩ ion. The Cu 2ϩ ions are exchanging sufficiently rapidly between peptide molecules to broaden all of the A␤ molecules with only 0.05 equivalents of Cu 2ϩ present. The rapid rate of exchange of the copper between peptides means that the broadening observed does not indicate quantitatively what fraction of Cu 2ϩ is bound. Unfortunately, due to the fast copper on/off rate, it is not possible to distinguish between high affinity copper coordination and a transient small fraction of the copper coordinating to, for example, a single histidine imidazole ring. It is therefore not clear from the NMR data whether the broadening of each histidine residue is due to a transient effect or a high affinity binding site. It is, however, clear that the lack of broadening of the Tyr 10 resonances indicates that it can be eliminated as a possible ligand in the coordination of copper.
Taking the 1 H NMR data together with CD data in both the UV and visible regions as well as the glycine competition studies of all of the A␤ analogues, we reveal a picture of the essential ligands for Cu 2ϩ coordination of the first mol eq of Cu 2ϩ ions to A␤. Our data strongly indicate the coordination of the N terminus of A␤ and His 13 , His 14 , and His 6 are also implicated, but not Tyr 10 .
Bridging Cu-His-Cu Dimer-It has previously been suggested that there is histidine bridging between two Cu 2ϩ ions in the A␤-(1-28) Cu 2ϩ complex (30). This work was based on the observation that the EPR spectra broaden with Cu 2ϩ addition (at pH 6.8) and assumed that this was due to electron coupling via a bridging histidine as is observed in EPR spectra of Cu 2 -superoxide dismutase (42). However, we have not observed broadening of EPR spectra with increasing copper addition. EPR spectra in which electron coupling is taking place are very temperature-dependent, and no such temperature dependence is observed in these spectra. Spectra have been obtained for A␤-(1-28) with 0.8 mol eq of Cu 2ϩ over a range of temperatures between 20 and 110 K, but line shapes are not affected by this change in temperature. We suggest that the broadening observed in this previous study (30) was simply due to through-space dipole coupling at high copper ion concentration rather than scalar through-bond coupling in a bridged complex for Cu-A␤-Cu.
Model Complex-A crystal structure of a model compound containing a dipeptide analogue of two adjacent histidine residues has been reported (43). In this compound, copper chelates to the ⑀N of both imidazole rings making a cis arrangement rather than trans coordination (90°rather than 180°) to the copper ion in a square-planar arrangement. Sterically, this type of ligand arrangement could facilitate the coordination of the N terminus and His 6 without any obvious restrictions. Interestingly, coordination to copper from adjacent histidine imidazole rings makes it impossible for main-chain coordination, via the amide or carboxyl, for either Val 12 , His 13 , His 14 , or Gln 15 . The absence of main-chain coordination is supported by the CD spectrum at neutral pH, which shows little chiral optical activity for d-d transitions. The lack of appreciable d-d transition CD bands suggests minimal vicinal effects (38), and this implies that there is no main-chain coordination at physiological pH and below, as previously discussed. A study of A␤ using Raman spectroscopy indicates Cu-⑀N at mildly acidic conditions and ␦N coordination at physiological pH and above (41). Imidazole coordination via the ⑀N makes main-chain coordination less likely than the more stable six-membered chelate ring between the ␦N and the His main-chain amide (44). This transition between ⑀N and ␦N coordination with pH is supported by our CD studies that indicate main-chain coordi-

DISCUSSION
With increasing interest in the role of Cu 2ϩ in AD (24,25,28,29,45), we have used a range of complementary spectroscopic techniques including CD, EPR, NMR, and fluorescence to study Cu 2ϩ binding to the A␤ peptide. In summary, Cu 2ϩ binds to the N terminus of A␤ and the three histidine imidazole rings but not Tyr 10 , as shown in Fig. 10. A␤ is essentially a random coil peptide in aqueous medium, whereas the amyloid fibrillar form of the peptide is rich in ␤-sheet (40). We were therefore interested in the possibility that Cu 2ϩ chelation could induce a ␤-sheet conformation in A␤. We have shown that the N-terminal tail folds back on itself to facilitate coordination via the N-terminal amino group as well as His 13 and His 14 . However, although copper chelation does induce ordering of the mainchain, CD studies of A␤-(1-28) indicate copper does not induce a typical ␤-sheet conformation. In addition, CD comparisons of A␤-(1-16) and A␤-(1-28) indicate that residues 17-28 are not directly affected by copper coordination in its soluble form.
We have shown that His 13 is a critical residue in coordinating the Cu 2ϩ ion. Interestingly, rat-A␤ lacks a histidine at position 13, and copper-induced aggregation of rat A␤ is not observed (11). Furthermore, wild-type rats do not exhibit ADlike pathology, although transgenic rats containing human A␤ can present with AD-like pathology. It has been shown that copper coordination has a profound influence on the solubility of A␤ (11). The 14 N-terminal amino acids of A␤ contain four carboxylic acid groups and four basic groups, the three histidine residues and the N-terminal amino group. Cu 2ϩ coordination may cause the loss of all four positive charges. Thus, copper binding will have a profound influence on the electrostatic surface of the A␤ peptide with a reduction in charged groups on A␤ and a change from an overall neutral charge to a highly negatively charged N-terminal tail. It is likely that this change in charged residues is the cause of the aggregation observed upon Cu 2ϩ binding.
We have shown that coordination of Cu 2ϩ to A␤ is highly pH-dependent. It is clear that there is a transition between two coordination geometries as the pH is raised. In the case of A␤-(1-28), the midpoint is pH 8.0. This observation agrees with a potentiometric study recently published, which suggests a transition between a 3N and 4N complex at pH ϳ8 (46). A study by Atwood et al. showed that Cu 2ϩ only induced aggregation of A␤ as the pH is lowered from physiological pH values to pH 6.8 (11). The change in coordination geometry, perhaps due to the loss of amide main-chain coordination at lower pH values, could be key to the pH dependence of the aggregation observed in the presence of Cu 2ϩ .
Key to the physiological significance for Cu 2ϩ binding to A␤ is its affinity. The concentration of extracellular Cu 2ϩ is typically 10 M in blood plasma, with extracellular levels of Cu 2ϩ reaching as high as 15 M (10). Cu 2ϩ dissociation constants reported here for A␤-(1-28) are 2 or 3 orders of magnitude lower than that of the extracellular copper concentration. This means that A␤ has sufficient affinity to bind copper at physiological levels of Cu 2ϩ . Previously reported affinities of A␤ for copper differ significantly. Initial studies by Atwood et al. put the Cu 2ϩ affinity at 4 and 0.3 M for A␤-(1-40) and A␤-(1-42), respectively (11). This value was revised using competitive metal capture analysis to an attomolar affinity (10 15 ) for A␤-(1-42) (24). Others have reported a more modest affinity of 2.0 Ϯ 0.8 M for A␤-(1-42) (31). Our K d of 10 -100 nM lies between these two values but is considerably closer to that determined by Garzon-Rodriguez et al. (31).
Copper binding curves obtained from EPR data strongly suggest copper saturation of A␤ at 2 mol eq of Cu 2ϩ . We are aware of only two previous reports of the stoichiometry of copper binding to A␤, one of which suggests a 1:1 stoichiometry (31). However, in agreement with our studies, Atwood et al. have reported copper saturation at ϳ2 mol eq for both A␤-(1-40) and A␤-(1-42), with a high affinity site at 0.5 mol eq of Cu 2ϩ (24). With no change in the appearance of spectra between 0.5 mol eq and 1 eq of Cu 2ϩ , both our EPR and CD data indicate that A␤-(1-28) does not form a dimeric Cu(A␤) 2 species under these conditions. In addition, from our EPR data we observe no evidence of A␤-(1-28) using bridged histidine coordination to form a dimeric species as previously suggested (30).
The very different affinities of Cu 2ϩ for A␤ described by Atwood et al. (24) could reflect the differences in the complexes formed. We suggest that the Cu 2ϩ binding affinities described by Garzon-Rodriguez et al. for A␤-(1-42) and ourselves for A␤-(1-28) reflect the soluble form of Cu-A␤. The mode of binding copper in solution may be different from that found in plaques. Atwood et al. (24) report a dimeric species Cu(A␤) 2 , whereas we see no evidence of a dimeric species forming for A␤-(1-28) in solution. Dimerization is a concentration-dependent phenomenon. The strong tendency for A␤-(1-42) to form aggregates may facilitate cross-linking and a Cu(A␤) 2 species. It is likely that soluble A␤-(1-42) initially forms a one-to-one complex as is seen for A␤- . The hydrophobic C-terminal residues 29 -42 do not directly coordinate to copper. Indeed, apart from Met 35 , which has been ruled out by a study using Raman spectroscopy (41) in which a S-Cu band was not observed, residues 29 -42 contain only aliphatic side chains, which do not possess lone pairs to facilitate Cu 2ϩ ion coordination. Rather, the C-terminal residues may promote aggregation that facilitates a cross-linked dimeric copper complex of A␤.
There have been relatively few structural studies of the Cu-A␤ complex. The paramagnetic copper center has hampered the determination of an NMR solution structure. Studies using Raman spectroscopy have suggested that Cu 2ϩ binds to the N␦ of the histidine imidazole and will coordinate via the N⑀ nitrogen under mildly acidic conditions (41,47). Our CD studies support this pH-dependent transition. In agreement with our CD and NMR data and glycine competition studies, the Raman  13 , and His 14 in a square-planar geometry. The model illustrates the dominant coordination geometry of the first mole equivalent of copper at physiological pH. Coordination at higher pH (pH 8.5 and above) probably involves main-chain amide coordination. A MolMol representation was obtained using distance geometry (DYANA), in which the metal-ligand bond lengths (1.9 Å) and angles (square-planar geometry) were constrained.
spectroscopic study suggests that all three histidines coordinate to copper. However, in contrast with our observations, Tyr 10 but not the N terminus was implicated in coordination (41). In agreement with our work, potentiometric studies of A␤ fragments (46) do implicate the N terminus in binding Cu(II).
Recent studies now present strong evidence to link Cu 2ϩ with AD. For example, Raman microscopy has provided direct evidence that copper and zinc are bound via the histidine imidazole rings of A␤ in isolated senile plaque cores (25). In addition, copper chelators can solubilize amyloid plaques and represent possible therapies for Alzheimer's disease, since copper chelators can inhibit amyloid accumulation in transgenic mice (26). Copper is also linked with the neurotoxicity of A␤ and free radical damage associated with Alzheimer's disease (16). Finally, trace amounts of copper in the diet have been shown to induce ␤-amyloid plaques and learning deficits in a rabbit model of Alzheimer's disease (28). To conclude, we believe that our studies have made a significant contribution to understanding Cu 2ϩ binding of A␤ in solution by identifying the key ligands coordinating the Cu 2ϩ ion and the effect of pH on the binding mode and structuring of the main chain of A␤.