Methylation of the Imidazole Side Chains of the Alzheimer Disease Amyloid-β Peptide Results in Abolition of Superoxide Dismutase-like Structures and Inhibition of Neurotoxicity*

The toxicity of the amyloid-β peptide (Aβ) is thought to be responsible for the neurodegeneration associated with Alzheimer disease. Generation of hydrogen peroxide has been implicated as a key step in the toxic pathway. Aβ coordinates the redox active metal ion Cu2+ to catalytically generate H2O2. Structural studies on the interaction of Aβ with Cu have suggested that the coordination sphere about the Cu2+ resembles the active site of superoxide dismutase 1. To investigate the potential role for such structures in the toxicity of Aβ, two novel Aβ40 peptides, Aβ40(HisτMe) and Aβ40(HisπMe), have been prepared, in which the histidine residues 6, 13, and 14 have been substituted with modified histidines where either the π- or τ-nitrogen of the imidazole side chain is methylated to prevent the formation of bridging histidine moieties. These modifications did not inhibit the ability of these peptides to form fibrils. However, the modified peptides were four times more effective at generating H2O2 than the native sequence. Despite the ability to generate more H2O2, these peptides were not neurotoxic. Whereas the modifications to the peptide altered the metal binding properties, they also inhibited the interaction between the peptides and cell surface membranes. This is consistent with the notion that Aβ-membrane interactions are important for neurotoxicity and that inhibiting these interactions has therapeutic potential.

Alzheimer disease (AD) 1 is characterized by the deposition of amyloid plaques; the major constituent of plaques is the amyloid-␤ peptide (A␤), which is proteolytically derived from the membrane-bound amyloid precursor protein (1)(2)(3). A␤ has been implicated as a causative agent in the progression of AD (4), and evidence is emerging supporting copper, iron, and redox silent zinc in the pathogenesis of AD (reviewed in Refs. 5 and 6). Two of the defining hallmarks of AD are peptide aggregation and the generation of reactive oxygen species, both of which are consequences of A␤ coordinating metal ions (5).
The importance of Zn 2ϩ in plaque formation has been emphasized by the finding that age-and female sex-related plaque formation in Tg2576 transgenic mice was greatly reduced upon the genetic ablation of the zinc transporter 3 protein, which is required for zinc transport into synaptic vesicles (7). The plaques represent "metallic sinks" because high concentrations of Cu (400 M), Zn (1 mM), and Fe (1 mM) have been found in amyloid deposits in AD-affected brains (8,9).
Results obtained in both transgenic animals and the clinic utilizing an orally available metal chelator, clioquinol, further highlight the importance of metal ions to AD pathogenesis. In Tg2576 mice, clioquinol treatment markedly decreased A␤ deposition and improved the general health and body weight (10). These results led to a pilot phase II clinical trial in which clioquinol inhibited cognitive decline and decreased plasma A␤42 levels in patients with moderate to severe AD (11).
A␤ possesses histidine residues at positions 6, 13, and 14 near its N terminus, a structural element that results in A␤ being capable of coordinating transition metal ions. A variety of spectroscopic studies have confirmed that the imidazole side chains of the histidine residues constitute the principal site(s) of metal coordination (12)(13)(14)(15).
Our NMR and EPR studies (13,14) in both aqueous solution and membrane-mimetic environments showed that the coordination of the metal ion is consistent with the three histidine residues all being involved in metal binding. EPR spectra recorded in PBS at Cu 2ϩ /peptide molar ratios of Ͼ0.3 have consistently demonstrated broadening. This can be explained by dipolar or exchange effects due to two Cu ions being within ϳ6 Å of each other, and the peptide was coordinating a second Cu 2ϩ atom in a highly cooperative manner at a site close to the initial binding site. We have hypothesized that this second site would result in a histidine residue bridging between the metal ions (distance between metals, ϳ6 Å); the resulting coordina-tion sphere would be reminiscent of that observed in superoxide dismutase 1 (13) (Fig. 1).
To investigate the role of the superoxide dismutase-like metal binding site on the properties of A␤, two novel A␤40 peptides were created, in which the histidine residues 6, 13, and 14 have been substituted with modified histidines where either theor -nitrogen of the imidazole side chain is methylated ( Fig. 1) to prevent the formation of bridging histidine moieties. The N-methylation of the His side chain does not affect the aromaticity of the imidazole ring; however, it does eliminate the tautomeric equilibrium, locking the structure into one of two forms. Methylation at these sites would not inhibit the initial metal binding site, but it does prevent the subsequent formation of a histidine bridge, inhibiting the formation of superoxide dismutase-like structures.
The acidic phospholipid spin label 1-palmitoyl-2-(16-doxyl stearoyl) phosphotidylserine (16NPS) was synthesized according to the method of Hubbell and McConnell (16) and checked for purity and to ensure that the number of spins/mol was Ͼ90% of the theoretical value (17). Synthetic palmitoyloleoyl phosphotidylcholine (POPC) was purchased from Sigma-Aldrich, and palmitoyloleoyl phosphotidylserine (POPS) was purchased from Avanti Polar Lipids.

Peptide Synthesis
All peptides were synthesized according to methods outlined previously (18). Due to the large degree of similarity between the peptides, large batches of resins were prepared and split at appropriate points in the synthetic pathway such that each peptide was prepared on a scale of 0.03 mmol.

Peptide Characterization
Peptide purification by RP-HPLC was performed on a Waters instrument controlled by Millenium software and equipped with photodiode array detection. The solvent system used throughout was buffer A (10 mM aqueous ammonium bicarbonate), buffer B (CH 3 CN). Semi-preparative RP-HPLC was performed using a Vydac 259VHP810 semi-preparative column (10 ϫ 250 mm, 8 m) (Hesperia) at a flow rate of 4 ml/min using a linear gradient of 10 -60% CH 3 CN over 30 min. Analytical RP-HPLC was performed using a Waters Symmetry 300 TM C4 analytical column (4.6 ϫ 150 mm, 3.5 m) at a flow rate of 1 ml/min using a linear gradient of 10 -60% CH 3 CN over 30 min and heated to 60°C. For matrix-assisted laser desorption ionization time-of-flight mass spectroscopy analysis, samples were mixed with ␣-cyano-4-hydroxycinnamic acid as matrix and examined using a Bruker BIFLEX instrument (Bremen, Germany) in linear mode at 19.5 kV or reflector mode at 19.5 kV, 20 kV gradient reflector.
Electrospray ionization mass spectra were obtained on a Quattro II triple quadrupole instrument (Micromass, Manchester, UK). Samples were introduced to the source via an HP1100 LC system (Hewlett Packard) equipped with solvent degasser, binary pumps, and autosampler. Samples were dissolved in 50:50 water:acetontrile with 1% formic acid to give a final concentration of ϳ10 g/ml. A carrier solvent of 50:50 (v/v) water:acetontrile was fed at 20 l/min to the mass spectroscopy probe. Data were acquired in continuum mode from 400 -2000 m/z in 6 -8-s scans, with an ion source at 100°C and sampling cone at 40 V. The acquired data were deconvoluted using the transform algorithm supplied with the Masslynx v3.2 software. Amino acid analysis of Fmocderivatized hydrolysates showed good correlation with expected values for all peptides.

Large Unilamellar Vesicle Preparation
LUVs of POPS and POPC (50:50) were prepared according to the method of Hope et al. (19) at a concentration of 50 mg/ml in 10 mM PBS (pH 6.8). Lipid solutions were passed through a 100-nm polycarbonate filter using Avanti Polar Lipids mini extruder apparatus at 37°C.

Circular Dichroism Spectroscopy
CD spectra were obtained using a JASCO model J-710 spectropolarimeter (Tokyo, Japan) controlled using the Jasco software provided. Spectra were background subtracted, adjusted for concentration, and smoothed using the algorithm provided with the system.
FIG. 1. Illustration of Cu 2؉ coordination by A␤ peptides. a, N3O coordination is supplied by the imidazole nitrogens of the three histidine side chains at positions 6, 13, and 14, and oxygen coordination is provided by the phenolic side chain of tyrosine; and b, the histidine side chain bridging between Cu 2ϩ atoms generating a site similar to the active site of superoxide dismutase. c, methylation of the imidazole side chain of the histidine. Incorporation of these modified forms of histidine into A␤ would allow the peptide to bind the initial metal ion but would prevent the formation of superoxide dismutase-like structures through histidine bridging two or more metal atoms. d, the sequence of A␤40, with the histidine residues that were modified shown in bold.
Aqueous TFE Gradients-Stock solutions of A␤40, A␤40(HisMe), and A␤40(HisMe) were prepared in HFIP, each at a concentration of ϳ1 mg/ml. Aliquots of the stock solutions were added to phosphate buffer (10 mM sodium phosphate, pH 7.4 or 5.5) containing 0, 10, 20, 40, 60, or 100% TFE (v/v), achieving a peptide concentration of between 10 and 12 mol, and allowed to equilibrate for 10 min. Samples were run in a 0.2-cm-path length cell at room temperature.
Peptides in SDS Micelle-containing Buffer-Stock solutions of A␤40, A␤40(HisMe), or A␤40(HisMe) were prepared in 0.1% aqueous trifluoroacetic acid/80% acetonitrile. To 300 l of sodium phosphate buffer at the required pH was added 20 l of 10% (w/v) aqueous SDS. Aliquots of the stock solutions were added to each solution to a concentration between 10 and 12 M and allowed to equilibrate for 10 min. Samples were run in 0.2-cm-path length cell at room temperature.
Peptides in LUV Solution-Samples of A␤40, A␤40(HisMe), or A␤40(HisMe) in LUVs prepared during EPR experiments were diluted 1:4 in PBS (pH 6.5) and run in a 0.2-cm-path length cell at room temperature.
Cu 2ϩ Binding Experiments-Stock solutions of A␤40, A␤40(HisMe), or A␤40(HisMe) were prepared in 0.1% aqueous trifluoroacetic acid/ 80% acetonitrile. Aliquots of the stock solutions were diluted in PBS (10 mM phosphate, pH 7.4). Cu(II)-glycine (0.1 mM stock) was added to give the desired ratio with the peptide and allowed to equilibrate for 10 min. Samples were run in a 0.2-cm-path length cell at room temperature.

Electron Microscopy
A␤40, A␤40(HisMe), and A␤40(HisMe) samples were prepared in 5 mM aqueous HCl (typically 10 g/ml) and aged by standing at 37°C for 5 days; samples were diluted 1:100 or 1:1000 prior to analysis. An aliquot of each sample was spotted onto carbon-coated single hole copper grids (ProSciTech, Kirwan, Australia), and the loaded grids were dried over agar gel. Precipitates were negatively stained with 0.5% uranyl acetate for 2 min and then washed several times with water. The samples were visualized on a Siemens ELMI-SKOP 102 electron microscope at 60 V.

Electron Paramagnetic Resonance
Continuous wave X-band EPR spectra were obtained using a Bruker ECS106 spectrometer equipped with a temperature controller and flowthrough liquid nitrogen cryostat. 65 Cu 2ϩ spectra were collected at 110 K from samples contained in 4-mm-inner diameter "Suprasil" quartz EPR tubes (Wilmad). In order to eliminate the possibility that any line broadening observed in the spectra might be due to freezing-induced localized concentrations of sample, 10% glycerol was added to aqueous peptide buffer solutions. Peptides as a freeze-dried powder were added to the desired concentration to a suspension of spin-labeled (16NPS; concentration, 1:100) LUVs in buffer, and the mixture was vortexed under N 2 for 10 min at 305 K in polypropylene tubes. 100 mM stock solutions of analytical reagent grade CuCl 2 were prepared, and the desired amount was added to the LUVs after addition of the peptide. Lipid samples (25 l) were contained in 0.8-mm-inner diameter quartz capillaries (Wilmad) and handled as described by Gordon and Curtain (17) to ensure reproducibility. Spectra were recorded at 305 K. Instrument parameters are as described previously (13,14).

High-performance Immobilized Metal Ion Chromatography
A 1-ml HiTrap TM Chelating Sepharose TM column (Amersham Biosciences) with iminodiacetic acid functionality was attached to a Waters HPLC instrument controlled by Millenium software equipped with photodiode array detection. All solutions were introduced to the column via the solvent delivery system at a flow rate of 1.5 ml/min or as an injected solution. The column was swelled in buffer C (10 mM sodium phosphate, 100 mM NaCl buffer, pH 7.4) treated with EDTA (50 mM) for 2 min and then washed with water (2 min). A solution of Cu(II)-glycine, Ni(II)-glycine, or Zn(II)-glycine (0.1 M, 2 ml) was then introduced as an injected solution. Excess metal solution was washed off with ultrafiltered deionized water (2 min), and the column was allowed to equilibrate with buffer C (5 min). Peptides were prepared as 1.0 mM solutions in 10% TFE, and single peptides (5 l) were introduced to the metalcharged column by injection. The peptide was eluted by a pH gradient of 0 -100% buffer D (10 mM sodium phosphate, 100 mM NaCl, pH 2.5) over 20 min at 1.5 ml/min. To determine the elution pH of the peptides, a curve of pH versus retention time was produced by measuring the pH of 1.0-ml fractions collected during metal-and sample-free runs. Significance between groups was tested at 95% using Bonferroni t test (SigmaStat, Version 2.03).

Hydrogen Peroxide Assay
Dichlorofluorescein diacetate (Molecular Probes) was dissolved in 100% dimethyl sulfoxide (5 mM; argon purged for 1 h at 20°C), deacetylated with 50% (v/v) 0.05 M NaOH for 30 min, and then neutralized (pH 7.5) to a final concentration of 1 mM. The reactions were carried out in Dulbecco's PBS, pH 7.5, in a 96-well plate (250 l/well) containing freshly prepared synthetic peptide (200 nM) copper-glycine (150 nM CuCl 2 and 900 nM glycine), and ascorbate (0, 1.1, 2.2, 4.4, 8.8, and 17.6 M) in the presence of deacylated dichlorofluorescein diacetate (100 M) and horseradish peroxidase (0.1 M) incubated at 37°C. There was a 4:3 molar excess of peptide to copper to ensure that there would be no free copper ions in the solution. Studies were performed on the day of reagent preparation. The fluorescence signal specific for H 2 O 2 was decreased in parallel samples co-incubated with catalase (40 units/ml; 100 nM). Fluorescence readings were recorded on a PerkinElmer Life Sciences LS55 plate reader (excitation, 485 nm; emission, 530 nm) against a standard curve of reagent-grade H 2 O 2 in PBS. All experiments were performed in triplicate.

Toxicity Assays
Primary Neuronal Cultures-Mouse embryonic cortices (E14) were isolated from BL6J/129sv mice, dissected free of meninges, and dissociated in 0.025% trypsin according published methods (20). Dissociated cells were plated in poly-L-lysine-coated sterile culture plates in plating medium (minimal essential medium, 10% fetal calf serum, and 5% horse serum) at 125,000 cells/cm 2 and maintained at 37°C in 5% CO 2 . After 2 h, the plating medium was replaced with Neurobasal growth medium (Invitrogen) containing B27 supplements (Invitrogen), 10 g/ml gentamicin, and 2 mM glutamine. This method resulted in cultures highly enriched for neurons (Ͼ95% purity) with minimal astrocyte and microglial contamination as determined by immunostaining of culture preparations using specific marker antibodies (data not shown). Both toxicity and peptide uptake experiments were done in fresh Neurobasal medium using B27 supplements lacking antioxidants.
Peptide Stock Solutions-A␤40, A␤40(HisMe), and A␤40(HisMe) were dissolved in 20 mM NaOH and diluted in PBS (5 mM sodium phosphate, 15 mM NaCl, pH 7.4) to give 200 M stock solutions of each peptide. Peptide stock solutions were sonicated in a water bath for 5 min and centrifuged for 5 min at high speed, and the supernatant was removed and aged by standing at 37°C for 5 days before being added to neurons at a concentration of 20 M in culture medium.
Cell Viability Assay-Cortical neurons were plated into 48-well plates and treated with aged peptide solutions for 4 days. The neurotoxicity of the peptides was challenged using catalase, a specific H 2 O 2 scavenger (2000 units of activity/ml medium; Sigma). Cell survival was monitored by phase-contrast microscopy, and cell viability was quantitated using the MTS assay as described previously (20). The data were normalized and calculated as a percentage of untreated vehicle control values. Data are shown as mean Ϯ S.E. Statistical comparisons between groups were done using Student's t test.

Peptide Binding and Cell Uptake
Cortical neurons were plated into 12-well plates and treated with 7.5 M aged peptide solutions for 5 h. An aliquot of the conditioned media for each peptide was kept to represent the concentration of added peptide. The cortical neurons in the wells were briefly washed in cold PBS or ice-cold 0.1 M sodium carbonate (carbonate wash), and the cells were scraped up from the wells, and proteins were extracted in 100 l of lysis buffer (0.05 M Tris, 0.5 M NaCl, 0.05 M MgCl 2 , and 1% Triton X-100). The untreated conditioned media and cell extracts were analyzed by Western blot techniques as described previously (20) to confirm the presence of peptide uptake by cells. The nitrocellulose immunoblots were probed with WO2, an anti-A␤ monoclonal antibody, detected using a peroxidase sheep anti-mouse immunoglobulin (Amersham Biosciences), and developed with ECL reagent kit (Amersham Biosciences). The chemiluminescent signals from immunoblots were quantitated using a charged-coupled device camera imaging system and GeneTool analysis software (GeneGnome; Syngene, Cambrige, UK).

RESULTS
Peptide Synthesis-The synthesis of all peptides was carried out using the optimized method reported previously (18), where standard continuous flow SPPS was carried out using 2% 1,8diazabicyclo[5.4.0]undec-7-ene/N,N-dimethylformamide as the deprotection reagent until residue Ser 8 , after which point the reagent was changed to 20% piperidine/N,NЈ-dimethylform-amide. The production of A␤40(HisMe) and A␤40(HisMe) required only a minor modification to the synthesis with the substitution of Fmoc-His(Trt)-OH for the His/Me equivalent. The modified histidine residue replaced the wild-type residue at each of the three positions His 6 , His 13 , and His 14 to generate two peptides, one in which all the histidine residues were methylated at the -nitrogen of imidazole side chain (A␤40(HisMe)), and the other in which all the histidine residues were methylated at the -nitrogen (A␤40(HisMe)). When A␤40 was co-injected into RP-HPLC with either of the modified histidine peptides (A␤40(HisMe) or A␤40(HisMe)), the peptides co-eluted, indicating that the contribution of the modified histidines to the overall hydrophobicity of the peptides was negligible.
CD spectroscopy was performed to determine whether the modifications significantly altered the structural properties of the peptides. All the peptides displayed a small amount of ␤-sheet structure after dilution from HFIP stock solutions into PBS buffer (Fig. 2). Increasing the TFE content of the buffer to 10% induced ␣-helical structures in all peptides. Similar helical structures were generated by all peptides in aqueous SDS micelle solutions (data not shown). These data indicate that methylation of the histidine imidazole side chains did not significantly affect the structural tendencies of the peptide.
Aggregation Properties of Modified His Peptides-To ascertain whether modification of the histidine residues affected the ability of the peptide to aggregate, fibril formation of the wild-type and modified histidine peptides was investigated using electron microscopy. Previous studies have indicated that fibril formation is more pronounced in low pH solutions (21,22). The electron micrographs of A␤40, A␤40(HisMe), and A␤40(HisMe) aged for 5 days in 5 mM aqueous HCl solution show negatively stained fibrils in all samples analyzed (Fig. 3). Qualitatively, the morphology of the A␤40(His/Me) peptides is similar to that of the wild-type A␤. Therefore, the addition of a methyl group represented a conservative modification to A␤ that has not significantly altered the structural properties of the peptide.
Metal Binding of the A␤40His/Me Analogues-HP-IMAC was employed to determine the relative affinities of wild-type A␤40 and the His/Me variants for different metal ions. The strength of peptide-metal ion interactions was assessed as a function of the pH required to elute the various peptides from the HiTrap TM iminodiacetic acid column charged with Cu 2ϩ , Zn 2ϩ , or Ni 2ϩ . All three peptides bound strongly to Cu 2ϩ immobilized on the HiTrap TM column (Fig. 4). Note that the pH at which the peptide was released from the Cu 2ϩ -charged column is approaching the pK a of the iminodiacetic acid ligand (3.0), and consequently release of the Cu 2ϩ ion-peptide complex from the column may also have occurred, indicating how avidly these peptides coordinated copper. Binding of the peptides to Zn 2ϩ and Ni 2ϩ was significantly less avid in all cases. The chemical non-equivalence between the HisMe and HisMe imidazole nitrogens was apparent in analysis of Zn 2ϩ and Ni 2ϩ binding. Metal binding by A␤(HisMe) did not differ significantly from that by wild-type A␤. The A␤(HisMe) peptide displayed a similar trend of strong Cu 2ϩ binding, weak Zn 2ϩ binding, and intermediate Ni 2ϩ binding, but the overall interactions were significantly weaker than the wild-type A␤ and A␤(HisMe) peptides, as judged by an increase in the pH required to elute the peptide from the column. The weaker metal-ligand interactions by A␤40(HisMe) were consistent with the differences in metal binding for each of the imidazole nitrogens (23).
The structural consequences of the differences in metal binding were studied by CD spectroscopy. Addition of metal ions to A␤ accelerates peptide aggregation (12,24,25) by increasing the ␤-sheet content of the peptides. The CD spectra for each of the peptides, A␤40, A␤40(HisMe), and A␤30(HisMe), were recorded in the presence of increasing concentrations of Cu 2ϩ .

FIG. 2.
Modification of the histidine residues has no effect on the secondary structure propensities of the A␤ peptides. TFE gradient CD spectra of (a) A␤40 wild-type, (b) A␤40(HisMe), and (c) A␤40(HisMe) at pH 7.4. All peptides display a small amount of ␤-sheet structure after dilution from HFIP stock solutions into aqueous buffer. Increasing TFE content of the buffer to 10% induced ␣-helical conformation in all peptides. ␣-Helical content increased with increasing TFE concentration for all peptides, and maximal ␣-helical content was achieved at 40 -60% TFE.
A␤40, which was random coil when initially dissolved in PBS, showed a gain in ␤-sheet content, as indicated by the change in the ellipticity at 222 nm, when the Cu 2ϩ concentration was increased (Fig. 5a). The A␤40(HisMe) behaved in a similar fashion to the wild-type peptide and was predominantly random coil in PBS, and as Cu 2ϩ was titrated into solution, the conformation of the peptide shifted to ␤-sheet (Fig. 5c). In contrast, A␤40(HisMe), although also initially random coil in PBS, did not display a corresponding shift to ␤-sheet structure when Cu 2ϩ was titrated into the solution (Fig. 5b).
EPR Studies-We have previously used EPR spectroscopy to characterize Cu 2ϩ binding to A␤ in PBS (13,14). At low concentrations of Cu 2ϩ , A␤ coordinated the metal ion with a 3N1O coordination sphere, giving an EPR spectrum with the intense g Ќ resonance typical of a type 2 Cu 2ϩ binding site (Fig. 6). The slight asymmetry of the g Ќ peak in this spectrum is probably an indication of a slight tilt in the axis of the Cu atom with respect to the perpendicular. The nitrogen ligands have been identified by NMR spectroscopy as the three His ligands at positions 6, 13, and 14 (13,15). The oxygen ligand was identified by Raman spectroscopy as the phenolic oxygen of Tyr 10 (12), and EPR studies of A␤ in which Tyr 10 has been replaced by an alanine residue support this finding (26).
The EPR spectra of Cu 2ϩ coordinated to A␤40, A␤40(HisMe) and A␤40(HisMe) for the Cu 2ϩ /peptide range of 0.3-0.75 mol fraction in PBS were collected (Fig. 6). The g ʈ and gЌ values of 2.29 and 2.05, respectively, from the A␤40 spectrum (Fig. 6, spectrum G) indicate that the peptide adopted the same 3N1O square planar conformation as reported previously (13). When the ratio of Cu 2ϩ to peptide was increased, there was significant line broadening (Fig. 6, spectrum H). We have attributed this to interactions between Cu 2ϩ ions occupying sites adjacent to the initial site occupied at lower Cu 2ϩ /peptide ratios. A plausible arrangement that would give rise to this effect is illustrated in Fig. 1b, where the His behaves as a bidentate imidazolate bridge and coordinates two Cu 2ϩ atoms in a Cu 2ϩ -His(N)-Cu 2ϩ arrangement, thereby bringing the two metal ions to within ϳ6 Å of each other. The arrangement of Cu 2ϩ ions with a bridging His residue mimics the active site of superoxide dismutase (27). Synthetic bidentate imidazolate bridge dimeric Cu complexes prepared as superoxide dismutase mimetics give rise to EPR spectra with similar broadening effects (28).
The EPR spectra of A␤40(HisMe) and A␤40(HisMe) indicate that both peptides coordinate Cu 2ϩ at a 0.3 metal/peptide ratio in a square planar 3N1O arrangement similar to wildtype A␤ (Fig. 6, spectra A and D); in both cases, the g ʈ and gЌ values are identical to those of the wild-type peptide. The His/Me derivatives were designed to modify the binding of metal ions to the A␤ peptide but still allow coordination of the first Cu 2ϩ atom. As the Cu 2ϩ /peptide ratio was increased, there was no line broadening due to exchange effects observed (Fig. 6, spectra B and C and spectra E and F), unlike that seen for A␤40. These data indicate that although the modified His residues can coordinate one metal ion, they are unable to coordinate a second metal ion and cannot form superoxide dismutase-like structures.
Catalytic Production of Hydrogen Peroxide-The ability of catalase to rescue A␤ toxicity implicates the generation of H 2 O 2 in the toxic pathway. We have reported previously that A␤ peptides in the presence of the redox active metal Cu and Fe are able to catalytically generate H 2 O 2 (29,30). The catalytic activity of A␤, A␤40(HisMe), and A␤40(HisMe) to produce H 2 O 2 was determined using a dichlorofluorescein diacetate assay (26). The V max and K m of H 2 O 2 production were determined by successively increasing the amount of the substrate ascorbic acid used in the reaction, and Lineweaver-Burk plots yielded straight lines (Fig. 7). The respective V max and K m values for A␤40, A␤40(HisMe), and A␤40(HisMe) are 14 Ϯ 1, 55 Ϯ 6, and 67 Ϯ 2 nM/min and 1.4 Ϯ 0.2, 3.3 Ϯ 0.6, and 2.8 Ϯ 0.4 M. Therefore, the mutant peptides were more efficient in producing H 2 O 2 than wild-type A␤.
Membrane Interactions-Our previous work utilizing nitroxide label lipids has established that, in the presence of Cu 2ϩ , the A␤ peptide can undergo a conformational shift to an ␣-helical structure and insert into negatively charged LUVs as an oligomeric assembly (13). The nitroxide label alone gives rise to an EPR spectrum (Fig. 8, spectrum a). The spectrum was characteristic of a paramagnetic label in which free motion was relatively unconstrained in the dimension perpendicular to the plane of the membrane, as seen in the unequal intensity of the three lines. Incubation of A␤40:Cu 2ϩ and A␤42:Cu 2ϩ with LUVs resulted in oligomeric insertion of the A␤ into the LUVs, as indicated by the appearance of an extra peak (arrow) in the spectrum of the labeled lipid (Fig. 8, spectra c and f). The extra FIG. 4. Wild-type A␤40 and His/Me peptides were passed over HP-IMAC to determine their relative affinities for metal ions. Peptides were eluted from immobilized metal ions using a pH gradient. An increase in the pH value at which the peptide is released indicates weaker binding to the IMAC column. The asterisk indicates a significant difference in the pH as compared with the wild-type peptide (p Ͻ 0.05).
FIG . 5. CD spectra of (a) A␤40 (wild-type), (b) A␤40(HisMe), and (c) A␤40(HisMe) in PBS buffer, pH 7.4. When initially dissolved, all the peptides exhibited random coil structure. The addition of 0.3, 0.6, or 1.2 equivalents of Cu 2ϩ induced a shift in conformation from random coil to ␤-sheet in the case of the wild-type peptide and A␤40(HisMe). A␤40(HisMe) did not show a similar shift in conformation with the addition of Cu 2ϩ . peak is characteristic of a nitroxide label with restricted movement as a result of peptide insertion into the membrane bilayer (31). Incubating labeled lipid with A␤42 minus Cu 2ϩ indicated the peptide did not insert into the LUVs (Fig. 8, spectrum e) and therefore requires the presence of the metal ion. The spectra of A␤40(HisMe) and A␤40(HisMe) in the presence of Cu 2ϩ (Fig.  8, spectra b and d, respectively) show no evidence of the motionally restricted lipid component. This result suggests that the metal-mediated insertion of an oligomeric assembly of A␤ is dependent on the formation of histidine bridges.
As reported previously, the metal-mediated lipid insertion of A␤ is accompanied by a structural transition to ␣-helical structure (13). In the absence of metal ions, the peptide assumes a ␤-sheet structure that is associated with the LUV surface. The CD spectra of A␤40 in the presence of Cu 2ϩ and LUVs show a strong ␣-helical conformation associated with the peptide inserting into the LUVs (Fig. 9) as reported previously (13,14). In contrast, the modified His peptides give rise to spectra with little or no signal, and given that a white precipitate was also observed at the bottom of the tube, this is consistent with the peptide aggregating and falling out of solution. Therefore, the peptides not only have failed to insert into the LUVs but also do not associate with the lipid surface.
Toxicity of the Modified His Peptides-An important issue is to determine whether the impaired ability of the analogue peptides to bind to metal ions and membranes affects their neurotoxic activity. Several studies have shown that the neurotoxicity of A␤ is enhanced in the presence of metal ions (29,30). Aged A␤40, A␤40(HisMe), or A␤40(HisMe) was applied to primary mouse cortical neuronal cultures, and the cell viability measured by the MTS assay. A␤40 is clearly toxic, with ϳ75% cell viability after treatment. The toxicity of the wildtype peptide was abolished by co-treatment of the cells with catalase, an enzyme that catalyzes the disproportionation of H 2 O 2 , consistent with A␤ toxicity being dependent upon the production of H 2 O 2 (30,32). Neither A␤40(HisMe) nor A␤40(HisMe) demonstrated any neurotoxic activity, indicat-ing that cross-bridging of the histidine residues with the metal ion is critical for toxicity (Fig. 10).
We have shown previously that the toxicity of A␤ is dependent on the peptide directly interacting with the lipid bilayer of the cell membrane (20). Given the apparent lack of interaction FIG. 6. EPR spectra of Cu 2؉ coordination by the A␤40 peptide and His/Me analogues. At low concentrations of copper, the wildtype A␤40 peptide produces a characteristic pattern indicative of N3O coordination (spectrum G). At higher copper concentrations, the spectrum shows considerable broadening, attributed to two Cu 2ϩ ions coming to within 6 Å of each other (spectrum G). The His/Me analogue peptides exhibit the same 3N1O coordination at low concentrations of copper, i.e. the initial binding site (spectra A and D, respectively), but at higher concentrations the broadening of the spectra (spectra B and C for A␤40(HisMe); spectra E and F for A␤40(HisMe)) was not observed, consistent with the analogue peptides being unable to form histidine bridges, thereby preventing the exchange phenomena between two more copper atoms.  16NPS (1:100). a, the vesicles alone in the absence of peptide. With the addition of A␤40, the spectrum was not altered (spectrum e), but when Cu 2ϩ was added to the solution, the spectrum was perturbed (spectrum f, indicated by the arrow). This spectral change is consistent with insertion of the peptide into the lipid vesicles. Unlike A␤, the His/Me peptides do not insert into the lipid bilayer in the presence of Cu 2ϩ , indicating that the multiple coordination of Cu 2ϩ ions and the formation of a multimer are required for membrane insertion (spectra b and d).
between the modified peptides and the LUVs, we examined the ability of the peptides to bind to the cortical neuronal cells. The amount of peptide bound was determined by Western blot. We were able to observe that A␤ bound to the neuronal cells, but we could not detect any evidence of the modified peptide binding to the cells, consistent with our results with the LUVs showing that the modifications to the histidine residues have substantially lowered the affinity of the peptides for membranes.
Peptide Binding and Cell Uptake-The amount of peptide bound to cortical cell extracts was quantitated by Western blotting, and the results are expressed as a percentage of the initial peptide concentration. Densitometry of the bands from Western blots indicated that for A␤40 total cell extract, 0.73 Ϯ 0.04% of peptide was bound, and following a carbonate wash, 0.52 Ϯ 0.07% was bound. There was no detectable membrane binding of either A␤40(HisMe) or A␤40(HisMe) to the cortical neurons, even after overnight exposure to autorad film. DISCUSSION Previous structural studies have suggested that one consequence of metal ions binding to A␤ is the formation of bridging imidazolate groups from the side chains of the histidine residues (12,13), generating structures that are similar to the active site of superoxide dismutase (Fig. 1). To investigate the possible significance of these structures, we have prepared A␤ peptides in which the imidazole side chains of the histidine residues have been methylated at eitheror -nitrogen (Fig.  1). Both the -nitrogen and -nitrogen positions are utilized as metal coordination sites by naturally occurring proteins/enzymes (33,34). N-Methylation locks the imidazole ring into one tautomeric conformation, thereby restricting each of the pyrrole/pyridine nitrogen environments to a fixed position in each of the isomers.
All the peptides co-eluted when injected onto a RP-HPLC column, indicating that the modifications do not significantly alter the hydrophobicity of the peptide. CD studies in aqueous/ TFE (Fig. 2) and SDS micelle solutions showed that the structural propensities of the peptides are not significantly altered. The modified peptides formed fibrillar aggregates (Fig. 3) with no significant difference in the incubation times required for fibril formation as compared with wild-type. These modifications are very conservative, and they do not significantly alter the physio-chemical properties of the peptide. A␤40(HisMe) consistently eluted off the IMAC column at a lower pH than A␤40(HisMe), indicating that metal coordination at the -nitrogen was of higher affinity (Fig. 4). The metal binding profile of A␤40(HisMe) is very similar to that of wild-type A␤, suggesting that the primary binding mode of wild-type A␤ is through the -nitrogen of the imidazole side chain of the histidine residues.
Cu 2ϩ coordination at His-nitrogen() is a feature of aggregated A␤, and this coordination mode is favored at slightly acidic pH (12). At neutral pH, deprotonation of a backbone amide occurs, and coordination at His-nitrogen() is common to the soluble aggregates of the peptide. Metal coordination at the -nitrogen is promoted by the availability of the amide backbone nitrogen to form a stable chelate ring (23). The CD spectra of Cu 2ϩ titrated into the various peptides are consistent with these assertions (Fig. 5). A conformational shift from random coil to a ␤-sheet structure is necessary for A␤ fibril formation, and this shift is promoted by the presence of metal ions that accelerate aggregation of the peptide from solution (35,36). Our CD data show that the addition of Cu 2ϩ to A␤40(HisMe) induced a ␤-sheet conformation. In contrast, the A␤40(HisMe) peptide coordinates metal at the -nitrogen and remained in conformational exchange with no defined secondary structure when Cu 2ϩ was titrated into solution.
The EPR data showed that when Cu 2ϩ was titrated into the PBS-buffered peptide solutions, the initial copper binding sites for all three peptides were similar and consistent with a square planar N3O coordination sphere (Fig. 6). The X-band EPR spectra could identify the atom type in the coordination sphere but could not distinguish between coordination at theor -nitrogen. When the Cu:peptide ratio was increased, there was significant broadening of the wild-type A␤ spectra (Fig. 6, spectrum H). This broadening is due to dipole-dipole interactions brought about by two or more Cu 2ϩ atoms being close together in space. One possible arrangement that is consistent with these data is the formation of an imidazolate bridge between Cu 2ϩ atoms. Metal ion binding to nitrogen() would reduce the pK a of nitrogen(), facilitating the coordination of a second metal ion (37). Broadening of the spectra occurs at concentrations below the addition of 1 equivalent of Cu (all peptide concentrations were determined by amino acid analysis, and the metal content was determined by inductively coupled plasma mass spectrometry), suggesting that coordination of the second Cu 2ϩ ion was a highly cooperative event, whereby occupation of the initial binding site gives rise to a second high affinity site (13). The lack of broadening by the modified peptides is consistent with the alterations affecting metal binding such that the Cu 2ϩ atoms cannot be coordinated adjacent to one another. Whereas the modified histidine residues can coordinate the initial Cu, the formation of subsequent cooperative bound sites involving an imidazolate bridge would be inhibited.
Both A␤ and the mutant peptides could generate H 2 O 2 catalytically in the presence of ascorbate as a reducing substrate (Fig. 7). Interestingly, the modified peptides were more efficient at producing H 2 O 2 compared with wild-type A␤. Although it is difficult to determine the precise reason for this difference, it may be that the oligomerization caused by the bridging imidazolate group has led to a more sterically hindered active FIG. 9. CD spectra of wild-type A␤40 and modified histidine peptides in LUVs in the presence of Cu 2؉ . The wild-type A␤40 adopts an ␣-helical conformation and is associated with insertion of the peptide into the LUVs. Peptides containing modified histidines give rise to spectra consistent with precipitation of the peptide from solution. site limiting access of the substrate. Alternatively, the formation of the bridging imidazolate has subtly altered the redox properties of the Cu at the active site such that the shuttling of valance states by Cu that is required for this reaction to proceed (26) is not as facile.
The addition of metal ions to wild-type A␤ is able to induce the formation of ␣-helical oligomeric assemblies that penetrate lipid bilayers (13,14). Methylation of the imidazole side chains of the histidine residues not only inhibited the formation of superoxide dismutase-like units but also inhibited peptide insertion (Fig. 8) and the associated structural transitions (Fig.  9). The formation of the bridging imidazolate groups was necessary for the oligomerization of the peptide into membranepenetrating units. We have shown previously that A␤ not inserted into the membrane is bound to the surface of the lipid, typically in a ␤-sheet structure (13,14). Interestingly, the modified peptides gave essentially no signal, and the data suggest the peptides precipitated from solution. We observed a similar phenomenon when we examined the binding of the peptides to cultured neuronal cells, where wild-type A␤ bound to the cells, but the modified peptides failed to bind. Our CD studies have indicated that under suitable conditions, the modified peptides can adopt both ␣-helical and ␤-sheet structures (Fig. 2). Therefore, the failure to bind to the lipid/cell membrane is not due to limits on the structural plasticity of the peptides. The data are consistent with the modified peptides having a lower affinity for membrane binding compared with wild-type A␤. The interactions of A␤ with lipid membranes are driven by electrostatic interactions (38), and it is possible that the methylation of the histidine residues has altered the electrostatic properties sufficiently to inhibit interactions with lipid membranes. The membrane interaction of spike glycoprotein G from vesicular stomatitis virus is one example of how the electrostatic properties of a histidine residue can mediate protein-membrane interactions (39).
Aged A␤40 was neurotoxic, whereas neither A␤40(HisMe) or A␤40(HisMe) exhibited any neurotoxic activity (Fig. 10). The neurotoxicity of A␤ is rescued by catalase and is therefore associated with the generation of H 2 O 2 ; in the presence of Cu 2ϩ , both modified peptides A␤40(HisMe) and A␤40(HisMe) were more efficient at generating H 2 O 2 than wild-type A␤. We have shown recently that the toxicity of A␤ can be increased by mutating Met 35 to Val (20). This increase in toxicity was associated with a higher affinity of the peptide for lipid membranes. We proposed that A␤ toxicity relied on site-specific effects. The data presented here are consistent with the model that A␤-mediated H 2 O 2 production per se is not sufficient for neurotoxicity. Our EPR, CD data, and cell binding studies indicate that A␤40(HisMe) and A␤40(HisMe) have a significantly lowered affinity for lipid membranes as compared with wild-type A␤. The interaction between A␤ and lipid membranes is driven by electrostatic interactions, and A␤ toxicity can be blocked by inhibiting these electrostatic interactions (40). Additionally, the rat/mouse form of A␤ that has His 13 mutated to an Arg has been shown to be nontoxic (29). Taken together, these data suggest that targeting the electrostatic interactions between A␤ and membranes that are mediated through the histidine residues of A␤ may have some therapeutic potential.