Capturing a Reactive State of Amyloid Aggregates

Background: Association of redox-active Cu2+ with aggregated Aβ in amyloid plaques has been linked with ROS and oxidative stress in AD. Results: Cu2+/Cu+-bound Aβ fibrils undergo a redox cycle reaction with ascorbate and oxygen to produce H2O2. Conclusion: Cu2+/Cu+ ions bound to histidines of Aβ fibril offer enzyme-like reaction centers. Significance: The first site-specific structural evidence is presented on Cu+-bound Aβ fibrils that generate ROS. The interaction of redox-active copper ions with misfolded amyloid β (Aβ) is linked to production of reactive oxygen species (ROS), which has been associated with oxidative stress and neuronal damages in Alzheimer disease. Despite intensive studies, it is still not conclusive how the interaction of Cu+/Cu2+ with Aβ aggregates leads to ROS production even at the in vitro level. In this study, we examined the interaction between Cu+/Cu2+ and Aβ fibrils by solid-state NMR (SSNMR) and other spectroscopic methods. Our photometric studies confirmed the production of ∼60 μm hydrogen peroxide (H2O2) from a solution of 20 μm Cu2+ ions in complex with Aβ(1–40) in fibrils ([Cu2+]/[Aβ] = 0.4) within 2 h of incubation after addition of biological reducing agent ascorbate at the physiological concentration (∼1 mm). Furthermore, SSNMR 1H T1 measurements demonstrated that during ROS production the conversion of paramagnetic Cu2+ into diamagnetic Cu+ occurs while the reactive Cu+ ions remain bound to the amyloid fibrils. The results also suggest that O2 is required for rapid recycling of Cu+ bound to Aβ back to Cu2+, which allows for continuous production of H2O2. Both 13C and 15N SSNMR results show that Cu+ coordinates to Aβ(1–40) fibrils primarily through the side chain Nδ of both His-13 and His-14, suggesting major rearrangements from the Cu2+ coordination via Nϵ in the redox cycle. 13C SSNMR chemical shift analysis suggests that the overall Aβ conformations are largely unaffected by Cu+ binding. These results present crucial site-specific evidence of how the full-length Aβ in amyloid fibrils offers catalytic Cu+ centers.

Progressive accumulation of amyloid plaques in the brain is a major pathogenic event that characterizes Alzheimer disease (AD) 2 (1,2), which is a multisymptom neural disorder. The primary components of the plaque deposits are amyloid fibrils of 40-and 42-residue amyloid-␤ (A␤) peptides (3). The AD plaques are reported to contain high concentrations of redoxactive metals such as Cu 2ϩ (ϳ0.4 mM) and Fe 3ϩ (ϳ1 mM) (4), and A␤ has been shown to bind these metal ions with high affinity (4,5). Accumulation of the redox-active metal ions by these amyloid aggregates has been proposed to promote formation of reactive oxygen species (ROS), which may lead to oxidative damages implicated as a common and major disruption observed in AD brains (6 -10). Indeed, previous in vitro experiments indicated that Cu 2ϩ ions in submicromolar concentrations along with A␤ peptide are sufficient for the generation of ROS (11)(12)(13)(14)(15)(16)(17). Recent studies demonstrated that the redox activity in AD lesions is inhibited by prior exposure of the tissue sections to copper and iron chelators (18 -20).
Oxidative stress caused by ROS is believed to be one of the key events in the pathogenesis of AD (14,21,22). Thus, the ROS production through the redox reaction of Cu 2ϩ -bound A␤ has been the subject of intense efforts while the Cu-A␤ complex has been a potential therapeutic target (23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33). Several reaction mechanisms have been proposed for the generation of ROS by Cu 2ϩ -A␤, yet the exact mechanism remains experimentally unknown (15). A compelling hypothesis from recent studies by mass spectroscopy and other techniques is that the Cu 2ϩ ion in complex with A␤ fibrils can be reduced to Cu ϩ by biological reducing agents, such as ascorbic acid, cholesterol, or dopamine (12), and that Cu ϩ may play a vital role in the production of ROS with subsequent oxidation of the His or Tyr residue on A␤ (6, 11, 15, 34 -37). It was recently proposed that during the production of ROS, the redox-active Cu 2ϩ ions bound to A␤ aggregates undergo a redox cycle (i.e. Cu 2ϩ 7 Cu ϩ ) in the presence of O 2 and biological reductant ascorbate, of which the reported concentration in the brain is in a range of 0.2-10 mM (16). Alternatively, it has been proposed that the oxidizable sulfur group on the side chain of Met-35 of A␤ may participate in the electron transfer redox reaction with Cu 2ϩ to produce ROS such as H 2 O 2 (34,36,38,39). Indeed, it has been shown that A␤ isolated from plaques contains oxidized Met-35 (39,40). However, the redox reaction mechanism and associated structural changes involving copper-bound A␤ aggregates in the production of ROS are not well understood due to lack of structural studies correlating the Cu ϩ /Cu 2ϩ -A␤ complex to the genera-tion of ROS. There have been intense efforts to study the structures of Cu 2ϩ -bound monomeric and aggregated A␤ by NMR, EPR, x-ray diffraction, and other methods (28,(31)(32)(33)(41)(42)(43)(44) to better understand the ROS production mechanism in AD at the molecular level. Despite the implicated importance of the less stable Cu ϩ state, there are only a handful of structural studies for Cu ϩ -bound A␤ (45)(46)(47)(48). Recent extended x-ray absorption fine structure spectroscopy and x-ray absorption spectroscopy studies showed that the N-terminal A␤ fragments A␤ (6 -14) and A␤ (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16) bind Cu ϩ in a linear two-coordination geometry possibly through two adjacent histidine residues (His-13 and His-14) (46). Additional NMR studies for A␤ (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16) indicated the involvement of all three histidine residues (His-6, His-13, and His- 14) in Cu ϩ binding via dynamic exchange between these ligands (47,48). These previous studies, however, focused mainly on short fragments of A␤ because of the challenges in capturing the transient Cu ϩ species with amyloid aggregates. Although Cu ϩ binding was examined for an SDS-stabilized oligomeric A␤  species in a recent extended x-ray absorption fine structure spectroscopy study, the analysis merely indicated the co-existence of various coordination motifs (41). Thus, very little structural information is currently available for Cu ϩ binding of physiologically relevant aggregated full-length A␤ such as A␤  or A␤ .
In this study, we examined the structures of Cu ϩ /Cu 2ϩbound A␤ fibrils in relation to its redox chemistry that is responsible for ROS generation. We show that the A␤(1-40) fibrils bound to Cu 2ϩ produce a major ROS, H 2 O 2 , under aerobic conditions in the presence of ascorbate. Redox conversions between Cu 2ϩ to Cu ϩ are clearly demonstrated by SSNMR, which is an increasingly powerful tool for structural analysis of insoluble protein assemblies (33, 49 -61) and protein complexes with metal ions (33,52,53,(62)(63)(64)(65). SSNMR results also suggest that Cu ϩ remains in complex with A␤ fibrils through His-13 and His-14 in the redox cycle. Narrow 15 N chemical shift line widths found for Cu ϩ -bound His-13 and His-14 indicate that Cu ϩ coordination is likely to make the His residues structurally more ordered in a specific tautomeric state. 13 C SSNMR analysis suggests that the overall conformations of A␤ in the hydrophobic core regions in fibrils are not altered by Cu ϩ binding or redox reaction. Our results demonstrate the first site-specific structural insights into the origin of ROS in AD through studying Cu ϩ binding to A␤(1-40) fibrils and its involvement in the redox cycle and resulting ROS production.
Synthesis and Purification of A␤  Peptide-A␤(1-40) peptide (NH 2 -DAEFRHDSGYEVHHQKLVFFAEDVGSNK-GAIIGLMVGGVV-COOH) was synthesized and purified as reported previously (56). Briefly, A␤(1-40) was synthesized using solid-phase peptide synthesis with standard Fmoc synthesis and cleavage protocols (54,56). The crude peptide was purified by HPLC using acetonitrile and water gradient with 0.1% trifluoroacetic acid. 13 C and 15 N labeling was introduced as described previously by incorporating Fmoc-protected uniformly 13 C-and 15 N-labeled amino acids at selected residues. The Fmoc protection of the uniformly 13 C-and 15 N-labeled amino acids was performed at the University of Illinois at Chicago Research Resource Center using the protocol of Fields et al. (66) Preparation of A␤ Amyloid Fibril-A solution of 1 mM A␤(1-40) was prepared by first dissolving the A␤(1-40) peptide in 10 mM NaOH by brief vortexing. Then the peptide solution was diluted to a final concentration of 100 M with 10 mM phosphate buffer containing 0.02% NaN 3 , sonicated for 30 s in an ice bath, and filtered through an Amicon ultracentrifugal filter (molecular mass cutoff at 50 kDa) at 3200 ϫ g and Ϫ5°C to remove pre-existing aggregates (55,56). The final pH of the A␤(1-40) solution was ϳ7.4. The concentration of A␤  was determined based on the UV-visible absorbance at ϭ 280 nm and ⑀ ϭ 1280 M Ϫ1 cm Ϫ1 (56). The A␤(1-40) fibrils were prepared by incubating the filtered solution at room temperature with constant agitation for 2 weeks. The fibril formation was monitored by ThT fluorescence assay (55).
Photometric Quantification of Hydrogen Peroxide-The quantification of H 2 O 2 was performed per the instructions given in the assay kit. A 48-l aliquot of 98 or 196 M A␤(1-40) fibrils with or without CuCl 2 (f Cu/A␤ ϭ 0.4 or 0.0) was added to each well, where f Cu/A␤ denotes a ratio of copper ion concentration with respect to the A␤ concentration in a monomeric unit. The peroxide detection reagent mixture of 50 l (46 l of assay buffer, 2 l of OxiRed probe solution, 2 l of horseradish peroxidase (HRP)) was immediately added to each well. Next, 2 l of 50 mM ascorbate solution at a neutral pH or H 2 O was added, gently mixed, and incubated at room temperature for 10 min before the first measurement. In the presence of HRP, the OxiRed probe reacts with H 2 O 2 to produce a product that can be detected at max ϭ 570 nm. The product was quantified by a Dynex microplate reader. The concentration of H 2 O 2 produced from Cu 2ϩ -A␤(1-40) fibrils and ascorbate was determined from the H 2 O 2 calibration curve. For control experiments on Cu 2ϩ -bound monomeric A␤ or A␤-free Cu 2ϩ solution, a 48-l aliquot of 98 M A␤ monomer solution with 40 M CuCl 2 or 40 M CuCl 2 solution was used. A 10 mM phosphate buffer was used to prepare the solutions at pH 7.4. The subsequent experiments were performed following the protocol described above. Three duplicates were performed for each trial, and the errors were estimated from the standard deviations.
Quantification of Reduced Cu ϩ by BCS-Cu ϩ ions associated with A␤ fibril were estimated by a photometric assay with BCS, a Cu ϩ -specific indicator that forms a Cu ϩ -BCS complex, which shows a new absorption at 483 nm. The assay was performed as described previously (67). For analysis of the supernatant, the A␤ aggregate was pelleted by centrifugation at 16,000 ϫ g for 10 min, and the resulting supernatant was analyzed for A␤ aggregates by ThT fluorescence measurements (55). The fibrils were completely removed with less than 1% remaining in solution as estimated by the ThT fluorescence. Then the supernatant was analyzed by BCS assay for the amount of Cu ϩ remaining in solution.
Electron Microscopy-A JEOL JEM-1220 transmission electron microscope at an accelerating voltage of 120 kV was used for the morphological analysis of A␤(1-40) fibrils with Cu 2ϩ and ascorbate. A 10-l aliquot of sample was spotted on a carbon-coated Formvar 200-mesh copper grid (Electron Microscopy Sciences, Hatfield, PA) for 30 s and dried with a tissue paper. The grid was negatively stained with ϳ10 l of 2% uranyl acetate solution for 1 min and again dried with a tissue paper. Finally, the grid was air-dried and analyzed at the University of Illinois at Chicago Research Resource Center.
Solid-state NMR-All SSNMR experiments were performed on Varian Infinity Plus and Bruker Avance III spectrometers with a home-built 2.5-mm MAS triple-resonance probe at 9.4 tesla ( 1 H frequency of 400.2 MHz). The spinning speed was set to 20,000 Ϯ 3 Hz throughout all experiments. Approximately 2.5-4.0 mg of labeled A␤(1-40) fibril sample was used in each experiment. 13 C chemical shifts were referenced to neat tetramethylsilane (TMS) using the secondary reference of adamantine 13 CH 2 signal at 38.48 ppm. In one-dimensional (1D) 13 C CP-MAS experiments, adiabatic CP transfer was used. During the CP period, the 13 C radio frequency (RF) field amplitude was linearly swept from 45 to 65 kHz during a contact time of 1.0 ms, while the 1 H RF amplitude was kept constant at 75 kHz. During the detection period, 1 H (TPPM) decoupling of 90 kHz was employed. The 1D spectra in Fig. 5 (A-C) were obtained with 512 scans each and were processed with Gaussian broadening of 100 Hz. The 1D 15 N CP-MAS spectra in Fig. 7 were measured with the 1 H spin-lock field strength of 1 /2 ϭ 50 kHz. During the CP period, the 15 N RF field amplitude was linearly swept from 25 to 35 kHz during a contact time of 900 s. The spectra were obtained with 15,360 scans each and were processed with Gaussian broadening of 2 ppm (Fig. 5, A-C) and 1 ppm (D-F), respectively.
The two-dimensional (2D) 15 N/ 13 C correlation spectra in Fig. 8 were collected by preparing the initial 15 N magnetization by adiabatic CP. After the t 1 evolution period, the real or imaginary component of 15 N magnetization was transferred to 13 C by 13 C-15 N CP, in which the difference of C and N was matched to the spinning frequency. The 13 C signals were detected during the t 2 period. For each t 1 point, 1496 scans of free-induction decay (FID) were accumulated with an acquisition period of 10 ms. A total of 70 complex t 1 points was recorded with a t 1 increment of 35.7 s. In the 2D 13 C/ 15 N correlation experiments, 1 H TPPM decoupling of 90 kHz was employed during the t 1 and t 2 periods, whereas continuous wave (cw) decoupling of the same amplitude was used during the 15 N-13 C CP period. The experimental time was 49 h each. The FID was processed by NMRPipe software (68) with Gaussian window function of 150 Hz applied along the t 1 and t 2 time domains for all the 2D 13 C/ 15 N experiments. For all the 2D 13 C/ 13 C correlation SSNMR data, a finite-pulse radio-frequency-driven recoupling (fpRFDR) pulse sequence with a mixing time of 1.6 ms and a 13 C -pulse width of 15 s were used (69). After the adiabatic CP and the following t 1 evolution period, a real or imaginary component of the magnetization was stored along the z axis. After the mixing period, the signal was collected in the acquisition period of 10 ms. For each t 1 point, 128 scans of the signals were accumulated. A total of 100 complex t 1 points were recorded with a t 1 increment of 38.1 s. The assignments of 13 C shifts for these residues are listed in supplemental Table S1.
Amyloid Fibril Samples for SSNMR Experiments-For SSNMR experiments, the A␤(1-40) fibrils were recovered by centrifuging the sample at 3200 ϫ g for 1.5 h (30 min at a time) at Ϫ5°C. The supernatant was removed, and the recovered fibrils were frozen in liquid N 2 and lyophilized. The lyophilized sample was then packed into a 2.5-mm MAS rotor for SSNMR experiments. The powder sample in the rotor was rehydrated by adding the supernatant buffer (2 l per mg of fibrils) and centrifuging down at 2000 ϫ g for 2 min. The rehydrated sample was incubated overnight at 4°C before data acquisition.
For the Cu 2ϩ -A␤(1-40) fibril sample, 2 l of a concentrated CuCl 2 solution (300 mM, pH 7.4) was added to 15 ml of the fibril solution (100 M in monomer equivalence) to give a final mole ratio f Cu/A␤ ϭ 0.4, and the sample was incubated at 4°C for 24 h with a brief vortexing to mix the sample. Then the Cu 2ϩ -bound A␤(1-40) fibrils were recovered by centrifugation at 3200 ϫ g for 1.5 h (30 min at a time) at Ϫ5°C. The supernatant was discarded, and the resulting fibril pellet was frozen in liquid N 2 , lyophilized, and packed into a 2.5-mm MAS rotor for SSNMR analysis and rehydrated as mentioned previously.
For the preparation of the Cu 2ϩ -A␤ fibril sample incubated with ascorbate, an ascorbate solution at 340 mM (0.7 l per mg of fibrils) was added to a 2.5-mm MAS rotor packed with ϳ3 mg of Cu 2ϩ -A␤(1-40) fibrils by centrifuging at 6000 ϫ g for 15 min. To avoid pH-induced chemical shift changes in the sample, the ascorbic acid solution was neutralized in equimolar NaOH before introducing into the rotor. The sample was then used for SSNMR experiments. For the 1D and 2D experiments in Figs. 5, 7, and 8, the mole ratio of A␤ fibrils to ascorbate is f asc/A␤ ϭ 1. The sample temperature was set to 12°C during the SSNMR experiments. For the 1 H T 1 measurements that require sample exposure to O 2 in the air in Fig. 3, the SSNMR rotor containing the Cu 2ϩ -A␤ fibrils and ascorbate with the rotor cap removed was placed in a centrifuge tube with a tiny hole on the cap and incubated at room temperature for ϳ14 h with gentle shaking at 100 rpm on a New Brunswick Scientific Excella E24 incubator shaker. In subsequent 1 H T 1 SSNMR measurements, the temperature was set to 24°C. The sample was exposed to air three times.
Solution NMR Experiments for Monomeric A␤-The 1 H solution NMR spectra in Fig. 10 were collected on a Bruker Avance 900-MHz spectrometer equipped with a cryoprobe at the Center of Structural Biology, University of Illinois at Chicago. The sample temperature was set to 10°C in all the experiments. Three samples were prepared for the regular 1

ROS Production by Cu 2ϩ -bound A␤ Fibril-
The production of ROS, such as H 2 O 2 from Cu 2ϩ -A␤ complexes, was observed when one or more physiological reducing agents are available (11,12). However, only a limited number of previous studies have reported the interaction between Cu 2ϩ and A␤ fibrils, and their proposed roles in the ROS production are somewhat contradictory (15,16). Thus, we first tested the generation of H 2 O 2 from the Cu 2ϩ -A␤(1-40) fibril complex in the presence of reducing agent ascorbate. Although our group and others have reported the presence of more cytotoxic diffusible aggregates for A␤ (3, 55, 56, 70 -72), we selected A␤(1-40) fibrils as our initial target because they are more stable and better characterized. In Fig. 1, we monitored the H 2 O 2 concentration by incubating a Cu 2ϩ -bound A␤(1-40) fibril sample (ϳ50 M A␤ and 20 M Cu 2ϩ ions) in the presence of 1 mM ascorbate (green bars) and in the absence of ascorbate (red bars) by photometric assay (see under "Materials and Methods"). The data clearly demonstrated the production of ϳ40 M H 2 O 2 within an hour under these conditions. However, in the absence of ascorbate, no detectable H 2 O 2 was produced. We previously confirmed 0.4 mol eq of Cu 2ϩ ions are strongly bound to A␤(1-40) in fibrils (33). Thus, there is no effect from excess or unbound free Cu 2ϩ in the production of H 2 O 2 . The ascorbate concentration at 1 mM was chosen based on the estimated physiological ascorbate concentration of 0.2-10 mM in the brain (16,73). The concentration of Cu 2ϩ is 1 order of magnitude lower than the copper ion concentration found in the AD plaques (0.4 mM) (4). Nevertheless, a submillimolar level of H 2 O 2 was generated within several hours from the Cu 2ϩ -A␤(1-40) fibrils/ascorbate solution. We also confirmed the generation of another destructive ROS, hydroxyl radicals (data not shown). Although it was found that H 2 O 2 is more efficiently generated by an equivalent concentration of free Cu 2ϩ ions without A␤ (78 Ϯ 6 M at 1 h incubation) or by Cu 2ϩ -bound A␤ monomer (78 Ϯ 3 M at 1 h) in the presence of 1 mM ascorbate, in the physiological environment a high concentration of free Cu 2ϩ ions or monomeric A␤ is not likely to be present. Less efficient H 2 O 2 productions by Cu 2ϩ -bound A␤ fibril may be attributed to limited accessibility of ascorbate due to bulky A␤ fibrils. We also confirmed that doubling the Cu ϩ and A␤ fibril concentrations ([A␤] ϳ100 M and [Cu 2ϩ ] ϭ 40 M) proportionally increased the H 2 O 2 production rate (85 Ϯ 1 M at 1 h). Thus, it is unquestionable that Cu ϩ -bound A␤ fibrils are promoting the ROS production under the near-physiological condition, which mimics the conditions for AD with plaque and copper ion accumulation. Taken together, the results have demonstrated H 2 O 2 generation from the Cu 2ϩ -bound A␤ aggregates, establishing relevance of this system used for the present structural studies with the ROS production, which was reported in AD.
In Situ NMR Detection of Redox Cycling for Cu ϩ /Cu 2ϩbound A␤  Fibrils-Mechanistically, the existence of redox cycling between Cu 2ϩ and Cu ϩ is strong evidence for the conversion of molecular oxygen into H 2 O 2 (11,12,16,74,75). Hence, we next examined the possibility of detecting the redox state of copper ions after addition of ascorbate by SSNMR. In general, it is difficult to examine the redox state of a metal ion when the ion is bound to an insoluble protein such as amyloid fibrils. SSNMR is one of very few methodologies that provide access to detailed molecular structures of amyloid fibrils (76). In addition, NMR relaxation parameters are very sensitive probes of paramagnetic ions such as Cu 2ϩ (77). By utilizing paramagnetic T 1 and T 2 relaxation enhancements by Cu 2ϩ in 1 H and 13  to Cu 2ϩ , diamagnetic Cu ϩ ions yield no paramagnetic relaxation enhancements. Thus, the redox reaction from Cu 2ϩ to Cu ϩ can be monitored by examining the NMR relaxation parameters (62). Fig. 2 shows the incubation time (t) dependence of 1 H longitudinal relaxation time (T 1 ) for Cu 2ϩ -bound A␤(1-40) fibrils after addition of ascorbate in a time span of 6 and 100 h. The 1 H T 1 relaxation time is monitored by the inversion recovery pulse sequence applied on the 1 H spins followed by a CP transfer to detect the 13 C signals under MAS conditions (78). The A␤(1-40) fibril sample used in the measurements is uniformly 13 C-, 15 N-labeled at His-14, Ile-32, Val-36, and Gly-37, and the sample was incubated with 0.4 mol eq of Cu 2ϩ . The initial 1 H T 1 value of hydrated Cu 2ϩ -A␤(1-40) fibrils without ascorbate is 96 Ϯ 3 ms. When 1 mol eq of ascorbate with respect to A␤ was added to Cu 2ϩ -A␤(1-40) fibrils, the 1 H T 1 value showed an ϳ6-fold increase to 576 Ϯ 13 ms after the incubation time (t) of 1.5 h (Fig. 2A). The T 1 value at t ϭ 1.5 h is comparable with that of Cu 2ϩ -free A␤ fibrils. Thus, the data indicate that paramagnetic Cu 2ϩ bound to amyloid fibrils is completely reduced to diamagnetic Cu ϩ by ascorbate. Interestingly, after 70 h the T 1 value gradually decreased to 60% of the highest value (Fig. 2B). It is likely that Cu ϩ is oxidized back to the more stable Cu 2ϩ when the concentration of ascorbate is gradually decreased. However, we found difficulties in quantitatively reproducing the re-oxidization, and the incubation time required for the 1 H T 1 decrease varied considerably from experiment to experiment.
After realizing the possibility that this re-oxidation reaction of Cu ϩ may require oxygen from air, we performed a modified experiment shown in Fig. 3 with A␤(1-40) isotope labeled at Val-18, Phe-19, Gly-29, and Ile-31. First, to speed up the redox reaction, we used 5 mol eq of ascorbate to A␤; the observed 1 H T 1 rapidly increased from 83 Ϯ 3 to 577 Ϯ 29 ms in 2 h. Second, in the modified experiment, after observing complete conversion of Cu 2ϩ to Cu ϩ that is reflected to the 1 H T 1 value, we made oxygen accessible to the sample by halting the 1 H T 1 measurement and opening the cap of the MAS rotor containing the A␤ sample. After overnight incubation at 24°C (Fig. 3, orange arrows), we resumed the 1 H T 1 measurements. This process was repeated three successive times. As indicated by the orange arrow in Fig. 3 at 5-13 h, the 1 H T 1 dropped considerably to 364 Ϯ 23 ms during the first incubation with oxygen; the shorter T 1 value by enhanced paramagnetic relaxation suggests that Cu ϩ was reoxidized to Cu 2ϩ in the presence of oxygen. At this point, it is likely that the excess ascorbate prevented the complete oxidization of Cu ϩ ions. Nevertheless, the 1 H T 1 value in Fig. 3 decreased much faster than the T 1 value for Fig. 2, which was measured for the sample without the exposure to air. Subsequent cycles of 1 H T 1 monitoring and exposure to air confirmed the rapid redox recycling between Cu 2ϩ and Cu ϩ in association with the amyloid fibrils. As 1 H T 1 of the amyloid fibrils is affected only when Cu 2ϩ is bound to A␤, the data clearly suggest that Cu 2ϩ is bound to A␤ throughout the redox cycle. We also confirmed the reduction from Cu 2ϩ to Cu ϩ at a physiological ascorbate concentration (1 mM) by a photometric assay using a Cu ϩ -selective indicator (Fig. 4). For a solution containing a suspended Cu 2ϩ -A␤ fibril complex (f Cu/A␤ ϭ 0.4) with an A␤ concentration of 100 M (in monomer equivalence), it took only 15 min for the Cu 2ϩ ions bound to A␤ to be nearly completely reduced to Cu ϩ (ϳ98%) (Fig. 4, red). The supernatant of the solution after centrifugation at 16,000 ϫ g for 10 min nearly completely removed Cu ϩ from the solution (Fig. 4,  green). This suggests that Cu ϩ ions are still bound to amyloid fibrils, which were pelleted down by centrifugation. Although previous studies implied Cu ϩ association with A␤, these studies used either fragments of A␤ or A␤ in monomeric forms rather than A␤ aggregates (45,46,67,79). Results from our SSNMR and photometric studies clearly demonstrate that Cu 2ϩ and Cu ϩ ions undergo redox cycling while bound to A␤ fibrils of full-length A␤ in the presence of ascorbate and oxygen for the first time. This also allows us to identify the Cu ϩ -bound A␤ fibrils as more reactive amyloid aggregates, which are most likely to be responsible for the ROS production. As will be discussed further below, our SSNMR approach offers a new avenue for in situ detection of binding modes for Cu ϩ as well as Cu 2ϩ ions with the aggregated proteins in the redox cycle.  Orange arrows show the periods during which the MAS rotor was opened, and the sample was exposed to the air. 13 C SSNMR-Next, we carried out 13 C SSNMR experiments to reveal the molecular level structural details and dynamics of Cu 2ϩ -A␤(1-40) fibrils during the redox cycle. Fig. 5, A and B, shows 13 C CP-MAS spectra of Cu 2ϩ -free and Cu 2ϩ -bound A␤(1-40) fibrils prepared with uniform 13 C-and 15 N-labeled amino acids at Val-12, Ala-21, Gly-33, and with selectively 13 C-labeled Met-35 (at the ⑀CH 3 position). Compared with the spectrum of the Cu 2ϩ -free sam-ple in Fig. 5A, Fig. 5B shows selective loss in signal intensities of 13 C ␣ , 13 C ␤ , and 13 C ␥ of Val-12 (blue arrows), whereas the chemical shifts or signal intensities of Ala-21 and Gly-33 are unaffected (33). The corresponding 1 H T 1 value of Cu 2ϩ -bound A␤(1-40) fibrils is 55 Ϯ 3 ms, which is much shorter than that of A␤(1-40) fibrils without Cu 2ϩ (370 Ϯ 18 ms). Selective quenching of Val-12 signals (Fig. 5B) and 1 H T 1 reduction are consistent with Cu 2ϩ binding at the side chains of His-13, His-14, and other residues of A␤ in fibrils (33). We then incubated the Cu 2ϩ -A␤ fibrils with 1 mol eq of ascorbate to A␤ at room temperature. After t ϭ 1.5 h, we collected a 13 C CP-MAS spectrum for this sample (Fig. 5C). Clearly, the 13 C signals of Val-12 recover to the original intensities (ϳ95%) of A␤(1-40) fibrils without Cu 2ϩ (Fig. 5C, red arrows). We attribute this to reduction of paramagnetic Cu 2ϩ bound to A␤ to diamagnetic Cu ϩ . The 1 H T 1 value at a 1.5-h incubation with ascorbate (351 Ϯ 14 ms) is also consistent with reduction of Cu 2ϩ to Cu ϩ . Based on the assumption that paramagnetic relaxation enhancement in 1 H T 1 is proportional to Cu 2ϩ concentration, then ϳ95% of Cu 2ϩ was converted to Cu ϩ . Although it is possible to explain this observation by the removal of Cu 2ϩ from A␤ due to chelation by ascorbate, the 13 C SSNMR data suggest that this is not the case as will be discussed below. The NMR data suggest that copper ions are subjected to redox cycling between Cu 2ϩ and Cu ϩ while remaining coordinated to the A␤ fibrils.

Redox Reaction and Structural Features of Cu ϩ -bound A␤ Fibrils Probed by
In addition to biological reductants such as ascorbate, the Met-35 residue of A␤ has been proposed to play a significant role in the redox chemistry of copper ions in this system as well (34). The methyl-13 C ⑀ of Met-35 has been assigned to a signal at   ϳ15 ppm (33), but the signal overlapping with methyl-13 C signals of Val-12 and Ala-21 (Fig. 5, A-C) does not allow for unambiguous determination of the paramagnetic influence from Cu 2ϩ . A minor peak at 40 ppm (indicated by # in Fig. 5) has also been assigned to the methyl-13 C of oxidized methionine (80). The fact that the intensity of this signal at 40 ppm did not change for Cu ϩ -bound fibrils suggests that no additional oxidation of Met-35 had occurred during the redox cycle (Fig. 5C). This small population of oxidized methionine probably resulted from the initial dissolution of A␤ peptide in a dilute sodium hydroxide solution (10 mM) or during the incubation period (34,80). A control experiment revealed significant oxidation of L-methionine after incubation with Cu 2ϩ and ascorbate (data not shown) (81). This indicates that the unique structural arrangements of A␤ fibrils may limit solvent accessibility of the Met-35 side chain (33) in the aggregated A␤ sample, thus preventing its involvement in the redox reaction.
Interestingly, the 13 C shifts of Val-12, Ala-21, Ala-30, Ile-32, Gly-33, Val-36, Gly-37, Gly-38, and Val-39 for the Cu ϩ -bound A␤ fibrils are essentially unaffected by Cu ϩ binding (see Fig. 5, A and C, and supplemental Table S1). 13 C chemical shifts are sensitive probes to conformational changes, and the results suggest that the overall A␤ structures in fibrils are largely unaffected by the reduction of Cu 2ϩ to Cu ϩ . Further analysis of these samples by 2D 13 C/ 13 C fpRFDR (69) experiment showed well resolved, nearly identical chemical shift cross-peaks (Ϯ0.2 ppm) (Fig. 5, D-F). Transmission electron microscopy indicated no morphological differences among these samples (Fig.  6). The lack of change in chemical shifts with indistinguishable transmission electron microscopy images suggests no significant alterations or degradation to the overall structure in the hydrophobic core of the fibril, although the local binding modes and conformations of the residues coordinated to ions may show structural differences. The results prove that amyloid fibrils of A␤ offer a robust architecture for the ROS production, which not only accumulates copper ions but also promotes copper-based reactions in a catalytic manner.
Histidine Coordination to Cu ϩ in A␤  Fibrils Probed by 1D 15 N and 2D 13 C/ 15 N SSNMR-To better understand the redox chemistry of copper-bound A␤ fibrils, we investigate the coordination modality of Cu ϩ on A␤ fibrils by SSNMR with focus on the coordination to His-13 and His-14 residues. As mentioned earlier, we reported that A␤(1-40) fibrils bind Cu 2ϩ via the side chains of His-13, His-14, and the carboxyl terminus of Val-40 and carboxyl side chains of Glu, based on SSNMR and molecular dynamic simulation studies (33). Previous NMR studies using a short A␤ fragment A␤(1-16) have proposed Cu ϩ binding at the histidine residues (i.e. His-6, His-13, and His- 14) in the N-terminal region of A␤ (47,48). However, no studies have demonstrated the interactions between Cu ϩ and A␤ fibrils. Herein, we conducted SSNMR analyses of copperbound and -free A␤(1-40) fibrils that are 13 C/ 15 N-enriched for His-13 and His-14. His-6 was not included in the analysis because the N-terminal region around His-6 of A␤(1-40) fibrils typically does not yield strong SSNMR signals presumably due to structural dynamics. Table 1 summarizes our chemical shift assignments for His-13 and His-14 from this study in comparison with those for published 13 C and 15 N shifts for histidine residues in various tautomeric states and Cu ϩ and Zn 2ϩ coordination states. We explain the interactions between Cu ϩ and A␤ suggested from the data below.
The imidazole side chain of histidine has two potential copper-binding sites, N ␦ and N ⑀ . To determine which nitrogen is coordinated to Cu ϩ , we performed the 1D 15 N CP-MAS (Fig. 7) and 2D 13 C/ 15 N correlation experiments (Fig. 8) on the same set of copper-free (A and D), Cu 2ϩ -bound (B and E), and Cu ϩbound 13 C-and 15 N-labeled (C and F) A␤ fibrils for His-13 (A-C) and His-14 (D-F). The 15 N chemical shifts of His side chains are highly dependent on the tautomeric state of the imidazole ring, which in turn depends on the pH of the sample (62,82,83). At the neutral pH used in this study, the deprotonated 15 N ␦ in the -tautomer and deprotonated 15 N ⑀ in the -tautomer both resonate at ϳ250 ppm, whereas protonated species 15 N ␦ H and 15 N ⑀ H resonate at 160 -190 ppm (see Table  1, b) (82,84). For copper-free A␤(1-40) fibrils (Fig. 7, A and D), the 15 N CP-MAS spectra show signals for His-13 at 228 and 172 ppm (Fig. 7A) and those for His-14 at 250 and 164 ppm (Fig. 7D) (82,83). Based on the 2D 13 C/ 15 N correlation spectra (Fig. 8, A  and D), the 15 N signals at 172 and 228 ppm for the His-13 sample are assigned to 15  are assigned by 2D 13 C/ 13 C correlation spectra (Fig. 9, A and D).
The lower signal intensity for 15 N ␦ is attributed to a lower CP efficiency of the nonprotonated species. Additional minor 15 N peaks are observed at 178 and 198 ppm for the His-14 sample (* in Fig. 7D), which are assigned, respectively, to 15 N ⑀ H and 15 N ␦ H for a bi-protonated His species as discussed below. The data indicate heterogeneous coordination environments for His-14 and possibly reflect hydrogen bonding from a neighboring residue. Upon Cu 2ϩ binding, the N ⑀ H signals for both labeled samples at ϳ164/172 ppm in the 15 N CP-MAS are quenched by ϳ40%, whereas the N ␦ signals at ϳ228/250 ppm are unaffected (Fig. 7, B and E); this observation suggests that Cu 2ϩ is coordinated to N ⑀ for both His-13 and His-14 (33). When Cu 2ϩ is reduced to Cu ϩ by ascorbate, the N ⑀ H signals at ϳ170 ppm are fully recovered, and new peaks at ϳ210 ppm emerge (Fig. 7, C  and F), which we assigned to Cu ϩ -bound 15 N ␦ based on 2D correlation spectra (Fig. 8, C and F) as described below.
In the 2D 13 C/ 15 N correlation spectra of copper-free A␤ fibrils (Fig. 8, A and D), the 15 N ⑀ H signal at 171.5 ppm for His-13 shows strong cross-peaks with directly connected C ␦ and C ⑀ , whereas the 15 N ␦ signal at 227.6 ppm displays cross-peaks only to C ␥ and C ⑀ . Thus, N ⑀ H (171.5 ppm) and N ␦ (227.6 ppm) are unambiguously assigned for His-13. Similarly, for His-14, we  13

C and 15 N chemical shifts observed for copper-free and Cu ؉ -bound His-13 and His-14 side chains of A␤(1-40) fibrils herein with published data on 13 C, 15 N chemical shifts of Cu ؉ -and Zn 2؉ -bound histidine residues in proteins and copper-free histidine in different tautomeric states
Part f shows the difference of 13 C ⑀ and 13 C ␦ shifts. The red number denotes the difference that is greater than 17 ppm.
* Data not known.   The cross-peaks corresponding to minor tautomeric forms of N ␦ and N ⑀ are shown in blue dashed circles. The disappearance of the N ␦ cross-peaks for the Cu 2ϩ -bound A␤ fibrils is due to the fact that the 15 N polarization is transferred to the paramagnetically quenched 13 C signals (B and E). The appearance of a new signal in C at 1 , 2 ϳ212, 136 ppm and in F at 1 , 2 ϳ208, 136 ppm after the incubation with ascorbate indicates Cu ϩ binding at the N ␦ (based on correlations to C ⑀ and C ␥ ) of the imidazole ring. The disappearance of the N ␦ cross-peaks for the Cu 2ϩ -bound A␤ fibrils is due to the fact that the 15 N polarization is transferred to the paramagnetically quenched 13 C signals (B and E). Binding of Cu ϩ at N ␦ of the imidazole ring is suggested by the appearance of new signals correlating 15 N ␦ to 13 C ␥ at ( 1 , 2 ) ϭ (212 and 136 ppm) in C and at (208 and 139 ppm) in F after incubation with ascorbate. The 2D 13 C/ 13 C correlation data were collected by the fpRFDR sequence (69) as discussed in Fig. 5, D-F. The 2D 15 N/ 13 C correlation data are reproduced from Fig. 8 for comparison. could identify correlations of 15 N ⑀ H at ϳ164 ppm with 13 C ⑀ at 136 ppm and 13 C ␦ at 114 ppm as well as the correlation of 15 N ␦ at ϳ250 ppm with 13 C ⑀ at 136 ppm. The spectrum for His-14 in Fig. 8D shows another sets of cross-peaks (blue dashed circles) involving 15 N signals indicated by the asterisks in Fig. 7D. These signals are attributed to the biprotonated form (82,83) of His-14 based on the 13 C and 15 N chemical shifts (see also Fig. 9).

His-14
Most of these resonances are quenched by Cu 2ϩ binding to His-13 and His-14 to A␤(1-40) fibrils and are in agreement with previous studies (Fig. 8, B and E) (33). Bi-protonated species for His-14 is negligible in Fig. 8E and suggests the replacement of a side chain NH proton with Cu 2ϩ .
Next, we examine the 2D 15 N/ 13 C correlation spectra of Cu ϩ -bound A␤ fibrils following incubation with ascorbate ( Fig.  8, C and F). In Fig. 8C, the spectrum for His-13 shows a notable cross-peak connecting the 15 N ␦ resonance at 212 ppm to 13 C ␥ at 136 ppm (red dotted square), whereas the 15 N ⑀ signal at 170 ppm correlates to 13 C ␦ at 116 ppm and 13 C ⑀ at 137 ppm. The 15 N ␦ resonance at 212 ppm is distinctively shifted from ϳ228 ppm and most likely assigned to N ␦ bound to Cu ϩ (N ␦ -Cu ϩ ; see Table 1). As the pH was adjusted carefully (see under "Materials and Methods"), the observed spectral changes are not attributed to the pH change. An additional minor cross-peak (blue circle) can be assigned to 15 N ␦ H/ 13 C ␥ or 15 N ␦ H/ 13 C ␦ when Cu ϩ is coordinated to N ⑀ . Comparable shifts have been observed for His-50 and His-128 of reduced copper-zinc superoxide dismutase (Table 1, b). We also observe similar patterns for His-14 to those above for His-13. The spectrum in Fig. 8F indicates correlations of 15 N ␦ -Cu ϩ at 208 ppm to 13 C ␥ at 139 ppm and 13 C ⑀ at 136 ppm (red dotted square) for His-14. There are also correlations of 15 N ⑀ H at 169 ppm to 13 C ␦ at 115 ppm and 13 C ⑀ at 136 ppm. The 15 N ␦ -Cu ϩ signals for His-14 sample show distinctively narrower peaks at 208 ppm unlike the broader multiple resonances observed for the copper-free sample in Fig. 8D. Similarly, the 13 C line widths for the His-13 and His-14 residues in the 2D 13 C/ 13 C spectra (Fig. 9, C and F) are narrower for the Cu ϩ -bound form than those for the copper-free form. These results suggest that Cu ϩ coordination to A␤ fibrils is likely to make the His residues more structurally ordered. The result shows an interesting contrast with 1 H solution NMR data for monomeric Cu ϩ -bound A␤(1-40), which display exchange broadening (Fig. 10) probably due to the transient nature of Cu ϩ binding to monomeric A␤.
The observed 15 N chemical shifts of ϳ210 ppm for 15 N ␦ -Cu ϩ are comparable with the reported 15 N ␦ shift of 220 -227 ppm for His-82 and His-125 in ba3 oxidase subunit II, for which N ␦ sites of the histidine residues are coordinated to Cu ϩ (see Table  1, c) (85,86). Furthermore, the 13 C ␦ chemical shift is known to be very sensitive to the protonation or Cu ϩ -coordination state of the neighboring N ⑀ . The 13 C ␦ shifts of His for the protonated and nonprotonated N ⑀ typically show ϳ115 and ϳ125 ppm, respectively (83,84). The 13 C ␦ shifts of Cu ϩ -A␤(1-40) fibrils observed at ϳ115 ppm (Fig. 9, C and F) indicate protonated N ⑀ , which is consistent with Cu ϩ coordination to N ␦ rather than N ⑀ unlike Cu 2ϩ . A recent statistical analysis of the 13 C ␦ and 13 C ⑀ chemical shifts of histidine coordinated to Zn 2ϩ (84) predicted that when the chemical shift difference between 13 C ⑀ and 13 C ␦ (⌬ ⑀␦ ) is Ͼ17 ppm, Zn 2ϩ coordination is likely to occur via the N ␦ (84). This empirical relationship of 13 C chemical shift data is also applicable to His coordinated to Cu ϩ for ba3 oxidase subunit II and superoxide dismutase (see Table 1) (85)(86)(87). Accordingly, the ⌬ ⑀␦ values of 21-22 ppm observed for His-13 and His-14 of Cu ϩ -bound A␤(1-40) fibrils (Table 1) suggest that Cu ϩ is most likely coordinated to the 15 N ␦ of His. Thus, the results indicate significant rearrangements of the copper-coordination structure through the redox conversion between Cu 2ϩ -A␤ and Cu ϩ -A␤. Weak cross-peaks observed at N ϭ ϳ170 ppm and C ϭ ϳ125 ppm in Fig. 8 (C and F) suggest the presence of minor species for which Cu ϩ may be coordinated to 15 N ⑀ . Such heterogeneous coordination modes have also been indicated in our previous study on the Cu 2ϩ -A␤ fibril complex, although additional data are needed to confirm assignments for the minor species. Taken together, the SSNMR results provide strong evidence that Cu ϩ ions preferentially coordinate to A␤ fibrils via the N ␦ of both His-13 and His-14 during the course of Cu 2ϩ /Cu ϩ redox cycling, which has been associated with the generation of ROS in AD.

DISCUSSION
Redox-active transition metals such as Cu 2ϩ and Fe 3ϩ have been implicated in the cause of extensive oxidative damages observed in the brains of AD patients (7,9,12,14,88). Researchers have exhaustively studied the interactions between these metal ions and AD-associated A␤ peptides (6, 7, 9, 11, 12, 31, 52, 88 -90). Nonetheless, given that these metal ions are isolated together with aggregated plaques/fibrils of A␤, a critical question remains unanswered. What is the role of these   peptide are assigned based on a previous report (30). As expected, the Cu 2ϩ ions broaden signals and in some peaks beyond detection in B. The amino acid residues affected by Cu 2ϩ include Phe, His, and Tyr. As reported previously by Hou and Zagorski (30), it is likely that Cu 2ϩ is bound to A␤ via the side chains of histidine (i.e. at His-6, His-13, and His-14) and some of the acidic amino acid residues on the N terminus. Reduction of Cu 2ϩ to Cu ϩ by ascorbate recovers a majority of the signals, but some remained broadened, especially the side chain protons of His residues, indicating the interactions between Cu ϩ and histidine amino acids of the A␤(1-40) peptide. The broadening is likely to be attributed to exchange broadening due to transient Cu ϩ binding. These results suggest A␤ binding of Cu ϩ at the histidine residues. Thus, the histidine residues of monomeric A␤  are involved in binding of both Cu ϩ and Cu 2ϩ . metal ions complexed to aggregated A␤ fibrils in regard to the oxidative stress in AD? In this study, we have shown the ability of Cu ϩ /Cu 2ϩ to form complexes with A␤(1-40) fibrils and generate harmful ROS such as H 2 O 2 and hydroxyl radicals as monitored by photometric assay and SSNMR. This study clearly illustrates the unique redox properties and structure of Cu ϩ -bound A␤ fibrils, which are likely to have relevance to the pathogenesis of AD. Quantitative analysis based on photometric assay suggests generation of submillimolar concentrations of H 2 O 2 by a 50-M A␤(1-40) fibril sample with a 0.4 mol eq of Cu 2ϩ in the presence of ascorbate. It was reported that the exposure to 50 -250 M H 2 O 2 for 24 h terminated ϳ50% of SH-SY5Y cells, which are a model cell system widely used for studying neural cell death (91,92). Furthermore, by utilizing various SSNMR experiments, we have shown that the electron transfer between Cu, ascorbate, and oxygen does occur and that both Cu 2ϩ and Cu ϩ remain bound to the fibrils during the redox reaction.
A generalized mechanism of the redox reaction entails initial reduction of Cu 2ϩ to Cu ϩ by ascorbate, resulting in H 2 O 2 formation; this is followed by an electron transfer from Cu ϩ to O 2 to oxidize back to Cu 2ϩ . The redox cycle indicated by our study is as shown in Reactions 1 and 2, Cu 2ϩ -A␤ ϩ Asc ¡ Cu ϩ -A␤ ϩ Oxd.Asc ϩ H ϩ

REACTION 1
Cu ϩ -A␤ ϩ H ϩ ϩ 1 ⁄2O 2 ¡ Cu 2ϩ -A␤ ϩ 1 ⁄2H 2 O 2 , REACTION 2 where Oxd.Asc denotes oxidized ascorbate. In the process, each O 2 can accept two electrons to generate H 2 O 2 . The presence of excess ascorbate and O 2 with Cu 2ϩ -A␤ can lead to a significant amount of locally accumulated H 2 O 2 and may result in oxidative damage of neurons. The requirements of cellular reductant and molecular oxygen for the ROS production imply that the diffusible A␤ aggregates, rather than immobile fibrils, may be a more efficient agent for catalyzing such reaction in terms of accessibility to cellular reducing agent(s). Such diffusible aggregates of A␤ have been reported to be more toxic to neural cells (56), although it is possible that the amyloid fibril of A␤ plays a major role in the ROS production at a late stage of AD where amyloid plaque is abundant. Furthermore, based on the SSNMR data, we propose that the Met-35 residue on A␤ is not involved in the redox cycle and that both His-13 and His-14 of A␤(1-40) fibrils can bind to Cu 2ϩ and Cu ϩ in different coordination modes through N ⑀ and N ␦ , respectively. As shown previously in our study on Cu 2ϩ -bound A␤ fibril (33), coordination of Cu ϩ ions to other potential ligands such as the C-terminal carboxyl group of Val-40 or Ala-42 may depend on conformations of A␤ in fibrils. Further structural studies will be required for elucidating fuller details of the coordination of Cu ϩ ions to A␤ fibril, which may influence the ROS production rates. Nevertheless, our previous study showed that Cu 2ϩ ions are likely to bind to His-13 and His-14 and other residues around the N terminus for different types of A␤(1-40) fibrils prepared by agitated and quiescent conditions (33). In summary, the results in this report indicate that A␤(1-40) fibrils become a strong catalyst that attracts copper ions and introduce cyclic redox reactions involving Cu ϩ /Cu 2ϩ ions in a continuous manner. Such reactions may contribute to the oxidative stress and a cascade of the downstream events in AD. Although additional studies are still needed, this study has presented the first detailed structural insight into the reactive Cu ϩ complex with A␤ fibrils with site specificity, which has not been achieved in many previous studies on copper ion-bound A␤ systems.