Distinct Effects of Zn2+, Cu2+, Fe3+, and Al3+ on Amyloid-β Stability, Oligomerization, and Aggregation

Abnormally high concentrations of Zn2+, Cu2+, and Fe3+ are present along with amyloid-β (Aβ) in the senile plaques in Alzheimer disease, where Al3+ is also detected. Aβ aggregation is the key pathogenic event in Alzheimer disease, where Aβ oligomers are the major culprits. The fundamental mechanism of these metal ions on Aβ remains elusive. Here, we employ 4,4′-Bis(1-anilinonaphthalene 8-sulfonate) and tyrosine fluorescence, CD, stopped flow fluorescence, guanidine hydrochloride denaturation, and photo-induced cross-linking to elucidate the effect of Zn2+, Cu2+, Fe3+, and Al3+ on Aβ at the early stage of the aggregation. Furthermore, thioflavin T assay, dot blotting, and transmission electron microscopy are utilized to examine Aβ aggregation. Our results show that Al3+ and Zn2+, but not Cu2+ and Fe3+, induce larger hydrophobic exposures of Aβ conformation, resulting in its significant destabilization at the early stage. The metal ion binding induces Aβ conformational changes with micromolar binding affinities and millisecond binding kinetics. Cu2+ and Zn2+ induce similar assembly of transiently appearing Aβ oligomers at the early state. During the aggregation, we found that Zn2+ exclusively promotes the annular protofibril formation without undergoing a nucleation process, whereas Cu2+ and Fe3+ inhibit fibril formation by prolonging the nucleation phases. Al3+ also inhibits fibril formation; however, the annular oligomers co-exist in the aggregation pathway. In conclusion, Zn2+, Cu2+, Fe3+, and Al3+ adopt distinct folding and aggregation mechanisms to affect Aβ, where Aβ destabilization promotes annular protofibril formation. Our study facilitates the understanding of annular Aβ oligomer formation upon metal ion binding.

Specific metal ions have been observed in the lesions of the disease. Analysis of the autopsy of AD patients shows abnormally high levels of specific metal ions present in the senile plaques (Cu 2ϩ , 25 g/g, ϳ393 M; Zn 2ϩ , 69 g/g, ϳ1055 M; and Fe 3ϩ , 52 g/g, ϳ940 M) (19,20). The levels of Cu 2ϩ , Zn 2ϩ , and Fe 3ϩ in the AD neuropil are also significantly elevated (19), where Zn 2ϩ is elevated from 346 to 786 M, Cu 2ϩ from 69 to 304 M, and Fe 3ϩ from 338 to 695 M (19,20). Al 3ϩ has also been detected in amyloid fibers in cores of the senile plaques (21). Moreover, the imbalance of cellular Zn 2ϩ and/or Cu 2ϩ homeostasis modulates AD pathology (22,23), and dietary Cu 2ϩ and Al 3ϩ are risk factors for AD (24 -25). These facts indicate that the elevation of the metal ions is relevant to AD pathology.
A␤ is able to bind to the metal ions (22,23). A␤ ion coordination, binding affinity, and induced aggregation have been studied intensively in various conditions; however, the results and mechanisms remain inconclusive. A␤-Cu 2ϩ coordination involves three intramolecular histidines (i.e. His-6, His-13, and His- 14) in A␤ (26 -28), with the fourth coordinate being either the amino group of the N terminus (29), an oxygen from Tyr-10 (30), or an oxygen from Glu-3 (29). The A␤-Zn 2ϩ complex is reportedly more complicated. A similar coordination with Cu 2ϩ has been proposed for Zn 2ϩ using the three histidines and the N terminus (31,32). Both intermolecular A␤-Cu 2ϩ and A␤-Zn 2ϩ coordination have been reported via histidine bridges (28, 30 -33). Fe 3ϩ has also been shown to interact with histidines (37). Dissociation constants for the binding affinity of ions and A␤ ranging from attomolar to 11 M for Cu 2ϩ and 2-300 M for Zn 2ϩ have been reported (38 -40).
The involvement of the metal ions with A␤ in AD and the potential development of metal chelating therapy indicate the importance of elucidating fundamental mechanisms of their effect on A␤. Here, we systematically examine the metal ion effects, especially of Zn 2ϩ , Cu 2ϩ , Fe 3ϩ , and Al 3ϩ , on early and aggregated stages of full-length A␤40. By employing different spectroscopic methods, including far-UV CD, tyrosine fluorescence, and 4,4Ј-Bis(1-anilinonaphthalene 8-sulfonate) (Bis-ANS) fluorescence, we monitor the conformational changes of A␤ and its ion binding affinity at the early stage. The binding kinetics and conformational stability are further examined by stopped flow machinery and guanidine hydrochloride denaturation. Photo-induced cross-linking of unmodified proteins (PICUP), thioflavin T (ThT) assay, dot blotting, and transmission electron microscopy (TEM) are also employed to monitor the oligomerization and fibrillization during aggregation. A mechanism for the metal ion effects on A␤ stability, oligomerization, and aggregation is then proposed.
A␤ Preparation-A␤40 peptide was synthesized using Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry and purified by reversed phase high performance liquid chromatography, as described previously (7). The molecular mass was identified by MALDI-TOF mass spectrometry (UltraFlex II; Bruker BioSciences, Billerica, MA). To prepare the A␤ stock, lyophilized peptide was freshly dissolved in Buffer A (10 mM Tris-HCl, pH 7.4) containing 8 M GdnHCl and refolded into Buffer A at a concentration of ϳ1 mg/ml. Then the stock was then centrifuged at 17,000 ϫ g at 4°C for 15 min. The supernatant was collected and quantified by absorbance at 280 nm (⑀ ϭ 1,280 cm Ϫ1 M Ϫ1 ) (55) and used as a stock solution to prepare A␤ at 25 M for all experiments.
Fluorescence Spectroscopy-Fluorescence emission spectra were obtained using a FluoroMax-3 spectrofluorometer (Horiba Jobin Yvon). The emission spectra of Bis-ANS at 5 M were collected from 450 to 550 nm with excitation at 400 nm.
The emission spectra of tyrosine were collected from 290 to 360 nm with excitation at 270 nm. Both fluorescent experiments were performed at 25°C. Temperature was controlled by a circulating water bath. The buffer backgrounds were subtracted.
Far-UV CD-Far-UV CD spectra were collected from 250 to 202 nm at 25°C using a Jasco J-815 spectropolarimeter (Jasco Inc., Easton, MD). A circular quartz cell with a path length of 1 mm was used. Six scans were performed and averaged for each condition.
Equilibrium Binding of Metal Ion and A␤ by Titration-The 25 M A␤ solution was titrated with different metal ion stock solutions: ZnCl 2 , CuCl 2 , and FeCl 3 at 2 mM or AlCl 3 at 20 mM. Less than 10% of the solution volume was increased after titration. Three different signals were collected, and the dilution factor was corrected. The monitored signals were 490 nm for Bis-ANS fluorescence, 305 nm for tyrosine fluorescence, and 216 nm for CD. The final titration signal was used as unity for normalization. The normalized data were plotted against the metal ion concentration. The amount of aluminum atom in the solution was confirmed by atomic absorption spectroscopy. Data fitting was performed by using an equation describing single protein ligand binding (56), where ␥ is the fraction of the observed signal changes representing the bound protein fraction, P t is the total A␤ concentration, L t is the total ligand concentration, and K d is the dissociation constant. In addition, the data were fitted to an equation describing one-protein and two-ligand binding, where ␥ is the fraction of the observed signal change, K 1 and K 2 are the association constants for the first and second ligand binding, and L is the free ligand concentration. Because the free ligand concentration cannot be faithfully determined in our experimental conditions, we used the total ligand concentration as L for the fitting. Stopped Flow Experiments for A␤ and Metal Ion Binding Kinetics-The kinetics of A␤ and metal ion binding was examined using a stopped flow module (Bio-Logic, Claix, France) attached to a Jasco J-815 spectropolarimeter. The stopped flow module was composed of a multimixer SFM-400, a motor power supply MPS-250, and a photomultiplier system PMS-250. A fluorescence cuvette, FC-15, with an optical path of 1.5 mm was used. Excitation at 270 nm was used to monitor the tyrosine fluorescence of A␤ at 12.5 M. A single mixing reaction using a volume ratio of 9:1 for the metal ions and A␤ was performed at 25°C. The final metal ion concentrations were 1.25 mM; hence the metal-to-A␤ ratio was 100 to 1. Here, A␤ was prepared in 100 mM Tris-HCl, pH 7.4, to avoid metal ion-induced acidity. In this buffer system, the reactions were in neutral pH. The flow rate was fixed at 11 ml/s, resulting in a dead time of 4.7 ms. The data were collected every 0.5 ms for the first 2 s, every 20 ms in the range of 2-60 s, and every 0.5 s for 60 -100 s. The data were fitted to multi-exponential equations using Bio-kine 32 V4.51 (Bio-Logic, Claix, France) with either single or double exponential equations with a linear base line, where a and b are the slope and offset for the linear base line, and c and k are the amplitudes and rate constants for the exponential phases, respectively.
GdnHCl Denaturation-The denaturation study was performed, as described previously (58), at 25°C. Briefly, the titration was performed by titrating an unfolded A␤ solution in Buffer A containing ϳ6.2 M GdnHCl into A␤ solution in Buffer A containing Ͻ0.1 M GdnHCl. The desired metal ion concentrations, 25 M of A␤, and 5 M of Bis-ANS were present in both solutions. The duration for each titration was ϳ30 s, and each set of denaturation experiment was less than 1 h. The Bis-ANS fluorescence emission at 490 nm was collected, averaged, and normalized. The normalized emission intensity versus GdnHCl concentration was plotted.
PICUP-The experiment was performed as described previously (59). Here, we prepared the A␤ stock by urea instead of GdnHCl to facilitate running of SDS-PAGE. Briefly, A␤ samples at 25 M prepared in Buffer A with different ion concentrations, as indicated, were immediately subjected to photoinduced cross-linking. A 90% A␤ solution was mixed with 5% each of 1 mM Tris(2,2Ј-bipyridyl)dichlororuthenium(II) and 20 mM ammonium persulfate. After mixing, the samples were exposed to a blue light LED in a closed chamber with a manual switch for 10 s. The cross-linking reaction was stopped by adding the SDS-PAGE sample buffer, and the samples were run on Tris-Tricine SDS-PAGE. All of the actions were performed without delay. The gel was subjected to Western blotting with anti-A␤ antibody 6E10 (Chemicon Inc., Billerica, MA) recognizing A␤ residues 1-17.
ThT Assay-A␤ (25 M) in Buffer A with different metal ion concentrations and 25 M ThT were incubated in an ELISA plate and monitored by a microplate reader (SpectraMax M5; Molecule Devices) at 25°C. The samples were in quiescence during the incubation, except for 10 s of mixing prior to the measurement. The ThT emission was measured at 485 nm, whereas excitation was at 442 nm. The signals were collected automatically every 1 h for the first 100 h, after which they were collected in a longer intervals.
Dot Blotting-To detect the oligomers during the aggregation, 2 l of the samples from the ThT assay were dotted onto a nitrocellulose membrane at the indicated incubation time and recognized using an anti-A␤ oligomer antibody, A11 (Invitrogen).
TEM-The samples were placed on glow-discharged, 400mesh Formvar carbon-coated copper grids (EMS Inc., Hatfield, PA) for 3 min, rinsed, and negatively stained with 2% uranyl acetate. The samples were examined with a Tecnai G2 Spirit TWIN TEM (FEI, Hillsboro, OR) with an accelerating voltage of 75 kV.
Sedimentation Assay-The aggregated products of 25 M Zn 2ϩ or 100 M Al 3ϩ and 25 M A␤ were incubated in Buffer A and subjected to centrifugation at 10,000 ϫ g at 4°C for 20 min. The supernatant and pellet were collected and analyzed by dot blotting using A11 and 6E10 antibodies and TEM. The pellet was resuspended in the original volume of Buffer A.

RESULTS
Al 3ϩ and Zn 2ϩ Increase Hydrophobic Exposure of A␤ Conformation, whereas Cu 2ϩ Decreases-To reveal the effect of metal ions on A␤, the ion-induced A␤ structural changes were monitored by three spectroscopic techniques: Bis-ANS fluorescence, tyrosine fluorescence, and far-UV CD spectra ( Fig. 1). Bis-ANS is able to report the hydrophobic clusters exposed on protein surfaces and to probe A␤ conformation at the early stage, where the binding sites are at the flanking regions of the protease-resistant segment of A␤ (58,60,61). The experiments were performed with A␤ at 25 M in the presence of Zn 2ϩ , Cu 2ϩ , Fe 3ϩ , and Al 3ϩ above their saturated concentrations, as determined by the metal ion binding experiment described below. The A␤ spectra with and without 200 M of Zn 2ϩ , Cu 2ϩ , and Fe 3ϩ and 500 M of Al 3ϩ are shown in Fig. 1A and supplemental Fig. S1. In the presence of Al 3ϩ and Zn 2ϩ , the Bis-ANS emission of A␤40 showed ϳ30and 15-fold enhancement, respectively, but decreased ϳ2.5-fold in the presence of Cu 2ϩ . The enhancement of the emission intensity indicates a larger extent of hydrophobic clusters exposed on the protein surface. Fe 3ϩ did not induce significant changes on the hydrophobic surfaces of A␤. Apart from the intensity difference, all of the spectra were blue-shifted from 500 nm to a range of ϳ480 -495 nm, indicating Bis-ANS encountering the hydrophobic environment.
In addition to extrinsic fluorescence, we used intrinsic tyrosine fluorescence at residue 10 of A␤ to report local conformational changes. In the absence of metal ions, the tyrosine emission had a maximum at ϳ302 nm while excited at 270 nm. We found that the tyrosine emissions were significantly quenched in the presence of Fe 3ϩ and Cu 2ϩ but enhanced in the presence of Al 3ϩ and Zn 2ϩ (Fig. 1B). The base lines of the spectra moved upward with concentration dependence to Al 3ϩ and Zn 2ϩ but not to Fe 3ϩ and Cu 2ϩ (supplemental Fig. S2). The increase in base line may be due to the formation of other unknown fluorescence species. Furthermore, far-UV CD spectra were employed to examine the secondary structure changes of A␤ (Fig. 1C). In the absence of the metal ions, a random coil-dominant spectrum was observed, as expected (62). However, all four metal ions further decreased ellipticity, especially with Zn 2ϩ and Al 3ϩ , revealing that the metal ions reduced the content of residual secondary structures existing in A␤ conformation. Together, our results showed that Zn 2ϩ and Al 3ϩ increased the hydrophobic exposed protein surfaces, induced higher tyrosine emission, and reduced most residual secondary structures of A␤. In contrast, Cu 2ϩ and Fe 3ϩ induced less or did not alter hydrophobic exposed surfaces, decreased tyrosine emission, and reduced residual secondary structures of A␤.
Ion Binding Stoichiometry and Affinity to A␤-To understand the conformational changes, binding stoichiometry, and affinity of A␤ and the metal ions, titration methods monitored by Bis-ANS fluorescence, tyrosine fluorescence, and far-UV CD were employed (Fig. 2). A␤ solutions at 25 M were used, and the signals were normalized as described under "Experimental Procedures." The fractions of A␤ bound with the metals were plotted against the metal ion to A␤ ratio. In general, the results showed that the three reporting signals changed coordinately but with some degree of differences. In the presence of Zn 2ϩ , three signals showed closed changes, and the data could be fitted to a single protein ligand interaction with the apparent dissociation constants, K d ϭ 3.9, 12, and 14 M for Bis-ANS, tyrosine, and CD signals, respectively. In the presence of Cu 2ϩ , both CD and tyrosine signals changed concomitantly and can be described by a single ligand binding with K d ϭ 6.3 and 4.9 M. However, the Bis-ANS signal was biphasic and delayed, suggesting that the hydrophobic exposed cluster of A␤ in the presence of Cu 2ϩ may be less sensitive to the changes of tyrosine environment and secondary structures.
In the presence of Fe 3ϩ , the three signals changed in a similar fashion; however, their standard deviations became larger. The data could be fitted to a single protein and ligand binding (data not shown) but were best fitted to the equation describing one-protein and two-ligand binding, assuming total ligand concentration as the free ligands. The K d1 and K d2 were 59 and 44 M for Bis-ANS, 185 and 55 M for tyrosine, and 606 and 8 M for CD signals. In the presence of Al 3ϩ , Bis-ANS signal changed faster than tyrosine signal, and tyrosine signal changed faster than CD signal. The Bis-ANS and tyrosine signals were best fitted to one-protein and twoligand binding (K d1 and K d2 were 108 and 61 M for Bis-ANS and 322 and 85 M for tyrosine), where CD signal was biphasic. Thus, the result suggested that in the presence of Al 3ϩ , the hydrophobic exposed surface was most sensitive to A␤ structural changes than the tyrosine environment, and the secondary structural changes were the least sensitive. The fitting suggested that the stoichiometry of Zn 2ϩ or Cu 2ϩ to A␤ is 1, and that of Fe 3ϩ and Al 3ϩ to A␤ may be 2. However, the exact concentration of Al 3ϩ ion in the solution was unknown because of the formation of aluminum hydroxide complexes in neural pH (63). Furthermore, higher concentration of Al 3ϩ (i.e. above 200 M in Buffer A)-induced acidity (supplemental Table S1).
Millisecond Binding Kinetics of the Metal Ions to A␤-The stopped flow apparatus monitoring total quantum yield of A␤ tyrosine fluorescence was employed to study the binding kinetics of the metal ions and A␤ (Fig. 3). The final metal ion concentration was 100 times more than A␤ concentration. The signal was monitored from the instrumental dead time, ϳ5 ms, up to 100 s. A␤ mixed with water had no signal change through time, indicating no conformational changes. In comparison with A␤ without ions, A␤s in the presence of Zn 2ϩ and Al 3ϩ contained more hydrophobic exposed protein surfaces but was more compact in Cu 2ϩ . A␤ in Fe 3ϩ did not differ significantly. B, tyrosine fluorescence spectra of A␤ with and without the metal ions. Zn 2ϩ and Al 3ϩ both increased the quantum yield of A␤, but Cu 2ϩ and Fe 3ϩ decreased the emission. C, far-UV CD spectra of A␤ with and without the metal ions. All of the ions altered the residual structures of A␤ to a lesser degree of the mean residue ellipticity.
When A␤ mixed with the metal ions, the signals showed exponential increases for Zn 2ϩ and Al 3ϩ , whereas the signal decayed exponentially for Fe 3ϩ . The signal for Cu 2ϩ had no significant change after 5 ms; however, the signal was already drastically decreased in the burst phase. The offset of Zn 2ϩ was elevated, similar to the base-line elevation in the steady state spectra (supplemental Fig. S2A). The offset changes for Al 3ϩ and Fe 3ϩ were smaller and may be neglected. Overall, the amplitude changes of kinetic signals were consistent with the changes observed in the steady state spectra (Fig.  1B), where Zn 2ϩ and Al 3ϩ increased the fluorescence, but Fe 3ϩ and Cu 2ϩ quenched the fluorescence.
The data were fitted to exponential equations with a linear base line as described under "Experimental Procedures," assuming pseudo-first-order reactions occurred. Fe 3ϩ and Al 3ϩ were best fitted to two exponential equations, whereas Zn 2ϩ was best fitted to a single exponential equation. The number of kinetic events reflected the binding stoichiometry mon-itored from the titration data. The rate constants and amplitudes are described in supplemental Table S2, and the linear time scale of the data is shown in supplemental Fig. S3. In contrast to Zn 2ϩ , Fe 3ϩ , and Al 3ϩ , Cu 2ϩ had nearly no signal change after the dead time, indicating that the events were completed in the burst phase. A very small increase observed at ϳ50 s can be attributed to diffusion. Therefore, the order of binding from the fastest to the slowest was Cu 2ϩ (submilliseconds) Ͼ Al 3ϩ (k 1 ϭ 6.067 s Ϫ1 , ϭ 0.114 s) Ͼ Fe 3ϩ (k 1 ϭ 3.455 s Ϫ1 , ϭ 0.20 s) Ͼ Zn 2ϩ (k 1 ϭ 0.124 s Ϫ1 , ϭ 5.553 s). The observed offset differences could be due to submillisecond binding events or environmental effects altering the tyrosine quantum yield. In the case of Zn 2ϩ , because the fitting precisely described the events from 0.005 to 100 s, and the elevation was also observed in the steady state spectra, we suggested that the elevation is not due to a binding event occurred in the burst phase.
Effects of the Metal Ions on the Stability of A␤-The possible effects of the metal ions to A␤ stability were then investigated. Although A␤ is considered to be intrinsically disordered (62), solution NMR and limited proteolysis studies have shown the existence of residual A␤ structures (64,65). Previously, we have shown that Bis-ANS fluorescence can distinguish "native" A␤, A␤ in native buffer, unfolded A␤, and A␤ in high concentrations of denaturant. The chemical denaturation coupled with data fitting is able to provide equilibrium folding mechanisms for A␤40 and A␤42 (58). Here, GdnHCl denaturation was employed to examine the stability changes of A␤ in the pres-ence of the metal ions (Fig. 4). In the native buffer, Bis-ANS binding to A␤ showed a significantly enhanced emission and blue-shifted spectrum in comparison with that in the presence of unfolded A␤ in 3 M GdnHCl (supplemental Fig. S4A). Tyrosine and CD spectra also showed different spectra between A␤ in native and unfolded conditions (supplemental Fig. S4, B and  C). The Bis-ANS signals of A␤ in the absence and presence of the metal ions were monitored in different GdnHCl concentrations, and the data were averaged and normalized. In the absence of the metal ions, a tilted pretransition, Ͻ0.5 M Gdn-HCl, followed by a single transition between 0.5 and 1.5 M Gdn-HCl, and a post-transition Ͼ1.5 M GdnHCl were observed, indicating A␤ adopted an apparent two-state equilibrium mechanism (N%U), where a "native" and an unfolded ensemble were present at the equilibrium (58). In the presence of 25 M Zn 2ϩ , 25 M Cu 2ϩ , 50 M Fe 3ϩ , or 100 M Al 3ϩ , the concentrations at the transition midpoints from the titration study, we found that Cu 2ϩ and Fe 3ϩ did not significantly alter the midpoint of the transition [GdnHCl]1 ⁄ 2 at ϳ1 M. On the contrary, Zn 2ϩ and Al 3ϩ shifted [GdnHCl]1 ⁄ 2 to ϳ0.4 and ϳ0.6 M, respectively. The pretransitions were lost in the presence of Zn 2ϩ and Al 3ϩ . The results demonstrated that Zn 2ϩ and Al 3ϩ drastically destabilized A␤. In addition, we performed reversed titration to titrate unfolded A␤ with A␤ in the native buffer. The denaturation was reversible in all conditions (supplemental Fig.  S5A). Alternatively, we monitored the denaturation by tyrosine fluorescence (supplemental Fig. S5B). The results also showed that A␤s in the presence of Zn 2ϩ and Al 3ϩ shifted the midpoints to lower GdnHCl concentrations. Other ions that do not bind to A␤, including Fe 2ϩ , Ca 2ϩ , Mg 2ϩ , and Na ϩ , had no effect on A␤ stability (data not shown).
Effect of Metal Ions on A␤ Oligomerization-We employed PICUP assay to detect transiently appearing A␤ oligomers in the presence of the metal ions at the early stage (Fig. 5). PICUP   was developed to photo-cross-link the short-lived metastable assembly in the fast equilibrium at the early stage (59). Here, we examined the effect in the presence of 25 M Zn 2ϩ , 25 M Cu 2ϩ , 50 M Fe 3ϩ , and 100 M Al 3ϩ . Before PICUP, A␤s were predominantly monomers, as examined by SDS-PAGE (data not shown). After PICUP, A␤ in the absence of the metal ions showed a dominant monomeric species and a minor dimeric species. No additional oligomer was induced in the presence of Fe 3ϩ and Al 3ϩ . However, in the presence of Zn 2ϩ and Cu 2ϩ , trimeric species were induced. Therefore, our results showed that Zn 2ϩ and Cu 2ϩ , but not Fe 3ϩ and Al 3ϩ , are capable of inducing higher assembly of A␤ oligomers that transiently appear at the early stage.
Effect of the Metal Ions on A␤ Oligomerization and Fibrillization-To investigate the kinetics of A␤ oligomerization and fibrillization, ThT assay was first applied to monitor the cross-␤-sheet formation during aggregation (Fig. 6). ThT is commonly used as a fluorescence probe in reporting the cross-␤ structures in amyloid fibrils (66). A␤ fibrillization is believed to undergo a nucleation-dependent pathway, where an oligomer nucleus is formed followed by an elongation event to mature fibrils (9). Five concentrations of the metal ions in different ranges were used, as indicated, for a better visualization of the effect. Without metal ions, A␤ followed the classic pattern of amyloid fibril formation to nucleate at a lag time of 20 h and elongated to reach a steady state after 60 h. In the presence of metal ions, concentration-dependent effects on A␤ fibrillization were observed. In the presence of Zn 2ϩ , the lag time was completely diminished, even in the lowest ion concentration (5 M), and ThT intensity increased readily without lag time (Fig.  6A). A plateau was seen after 20 h with a lower ThT intensity in comparison with A␤ alone. Apart from Zn 2ϩ , the other three metal ions had similar inhibition effects, albeit occurring at different concentrations of ions. At 5 M, Cu 2ϩ prolonged the lag phase significantly to more than 60 h and decreased ThT fluorescence (Fig. 6B). At 50 M, Cu 2ϩ completely abolished the aggregation. Fe 3ϩ and Al 3ϩ above 50 M both prolonged the lag phase to ϳ40 h, where Al 3ϩ has a strong inhibition effect above 250 M (Fig. 6, C and D). A weaker inhibition effect from Al 3ϩ was observed when the pH was neutralized by performing the experiment in Tris buffer at pH 7.85 (supplemental Fig. S6A). Thus, the strong inhibition may be primarily due to the acidic pH.
The corresponding oligomer appearance was further examined by dot blotting probed by anti-amyloid oligomer antibody, A11. A11 is an amyloid oligomer-specific antibody recognizing common epitopes in various amyloid proteins (13). A␤ incubated with different metal ion concentrations were spotted on the nitrocellulose membrane at different times during the aggregation. In the absence of metal ions, no significant A11positive oligomer appeared. In the presence of Zn 2ϩ , starting from 25 to 200 M, A11-positive signal was significantly enhanced and appeared at the first time point (Fig. 6A). For Cu 2ϩ and Fe 3ϩ , no obvious A11-positive signal was detected (Fig. 6, B and C). Al 3ϩ induced A11-positive signals in a timedependent manner; however, the signals disappeared in the presence of 500 M Al 3ϩ (Fig. 6D). The disappearance of A11 signals was not caused by the acidic pH because the signal was also significantly weakened at higher concentrations of Al 3ϩ when the solution was kept neutral (supplemental Fig. 6B).
Morphology of the Aggregates-We further employed TEM to observe the morphology of the A␤ aggregates in the presence of various metal ion concentrations (Fig. 7). In the absence of the metal ions, A␤ formed large amounts of mature fibrils. In the presence of Zn 2ϩ , no fibrils were found. Instead, massive amounts of heterogeneous oligomers mixed with several amorphous aggregates were observed. In contrast, with increased Cu 2ϩ concentration, the appearance of fibrils decreased; however, nonfibril, amorphous aggregates increased. Some short and fragmented fibrils were also observed at 25 M of Cu 2ϩ . A similar phenomenon was seen in the presence of higher Fe 3ϩ concentration, although without observing the short fibrils. A decrease in fibrils was also observed in the presence of Al 3ϩ . However, at 100 M Al 3ϩ , fibrils and oligomers were observed to co-exist. At 500 M Al 3ϩ , the solution became acidic, and the aggregates were mainly amorphous.
We further examined the secondary structural content of the aggregates by far-UV CD spectra (supplemental Fig. S7). The spectra of A␤ with 50 M of Fe 3ϩ showed similar ␤-sheet-dominant spectra to A␤ fibrils alone, indicating fibril formation in 50 M Fe 3ϩ . In the presence of 100 M Al 3ϩ , the ellipticity was reduced. The aggregates formed in 25 M of Cu 2ϩ and Zn 2ϩ were significantly different from ␤-sheet-rich fibrils, where the spectra of A␤ in the presence of Zn 2ϩ had a helix-like double minima spectra. However, the minima were not at the classic wavelength for ␣-helices. This result is consistent with the data obtained from ThT assay and TEM images.
To further characterize the Zn 2ϩ -and Al 3ϩ -induced oligomers, sedimentation was employed to separate the aggregates in soluble and insoluble fractions. The fractions were then analyzed by dot blotting and TEM (Fig. 8). At 10,000 ϫ g for 20 min, the A11 signals for both Zn 2ϩ and Al 3ϩ appeared mostly in the pellets, whereas half of the amount of the total protein remained in the supernatants for Zn 2ϩ and little protein for Al 3ϩ as demonstrated by 6E10 immunoblotting (Fig. 8A). TEM analysis (Fig. 8B) showed large amounts of heterogeneous, ringshaped, and pore-like oligomers in the pellet of A␤-Zn 2ϩ aggregates with an average diameter of ϳ23 nm and a pore size of ϳ12 nm. Weakly stained small oligomers that were difficult to count were also present. Fewer oligomers remained in the supernatant, and the species appeared to have a rougher surface. Close examination showed no obvious spherical oligomers, only mostly pore-like conformers. In the aggregates of A␤-Al 3ϩ , we found fibrils and oligomers co-existed in the pellet. Most oligomers were small (Ͻ10 nm) and had smooth surface; however, some large oligomers (ϳ25 nm) without pores were also present, especially in close proximity with the fibrils. A few rugged oligomers were observed in the supernatant, but no fibrils. Similar results were observed for dot blotting and TEM by examining the aggregates of A␤ in the presence of 250 M Al 3ϩ starting form Tris buffer, pH 7.85 (supplemental Fig.  S6, C and D). Pore-like oligomers and little curvilinear protofibrils were mostly observed in the supernatant, whereas fibrils and oligomers were observed to co-exist in the pellet. Those ring-shaped, pore-like oligomers most resembled the annular protofibrils described in the literature (67).

DISCUSSION
To understand the involvement of specific metal ions in A␤, we examined the influence of the metal ions on A␤ initial conformation and stability and the mechanism of the effect of these factors on its aggregation process. All four ions, i.e. Zn 2ϩ , Cu 2ϩ , Fe 3ϩ , and Al 3ϩ , reduced residual secondary structures of A␤ upon formation of metal-A␤ complex, but only Zn 2ϩ and Al 3ϩ induced larger exposure of the hydrophobic protein surfaces. Upon ion binding, a structural change was observed, and the binding occurred in submillisecond or millisecond time scale, in which Cu 2ϩ binds faster than Al 3ϩ , Fe 3ϩ , and Zn 2ϩ . The binding stoichiometries were suggested to be 1 for Zn 2ϩ and Cu 2ϩ and 2 for Fe 3ϩ and Al 3ϩ . GdnHCl denaturation showed that Zn 2ϩ and Al 3ϩ , but not Cu 2ϩ and Fe 3ϩ , destabilized A␤. Further examination on the metal ion effect on the aggregation process revealed that Zn 2ϩ preferentially facilitates annular protofibril formation without a nucleation phase, whereas other metal ions inhibit fibril formation, likely by partitioning A␤ into fibril or amorphous aggregate/protofibril formation pathways. The results are consistent with previous reports on the metal ion effects on A␤ aggregation, where Zn 2ϩ and Cu 2ϩ accelerate A␤ aggregation to form nonfibrillar aggregates but not ThT-positive fibrils (41)(42)(43)(44)(45)(46)(47)(48), and Al 3ϩ promotes A␤42 fibrils or oligomers formation (49,68).
The fact that both Zn 2ϩ and Al 3ϩ induced more hydrophobic exposure and destabilization of initial A␤ conformation, resulting in annular protofibrils formation indicates that A␤ conformation and stability are key factors in the formation of specific aggregates such as annular protofibrils. Although Cu 2ϩ induced slightly more compact A␤ conformation, the stability was not significantly different, suggesting that the Cu 2ϩ -induced retardation of fibril formation may not be due to the stability change of A␤. For A␤ aggregation in the presence of Cu 2ϩ , Fe 3ϩ , and Al 3ϩ , the partition between amorphous aggregates/protofibrils and fibrils seems to correlate with to their binding affinities, where the inhibition is more profound in higher concentration of Fe 3ϩ and Al 3ϩ . Although Zn 2ϩ and Cu 2ϩ induced similar trends of transiently appearing A␤ oligomers at the early stage, their aggregations were distinct, as reflected by ThT intensity, A11 dot blotting, and aggregate morphology. Our results demonstrated that the four metal ions adopt different mechanisms to influence A␤, yet similarities in partitioning pathways or stability influence on annular protofibril formation are present. A␤ is intrinsically disordered; however, residual structures that contain a turn connected by two loose ␤-strands are revealed by solution NMR (64) and limited proteolysis (65). The structure of the N terminus of A␤ from residues 1 to 10 is considered to be flexible and disordered. The region is not incorporated in the structural assembly of fibrils (69) and oligomers (70). The two histidinyl residues, His-13 and His-14, involved in the metal ion coordination are located at the edge of the first ␤-strand in fibrils (69) and are involved in a solvent accessible turn formed between His-13 to Gln-15 in discshaped A␤42 oligomers (70). Our results showing A␤ structural changes concomitantly upon metal ion binding demonstrated that the ion binding at the N terminus is able to affect the overall residual conformation of A␤. Several familial AD mutants containing single mutations at the N terminus such as English mutant, H6R (71), and Tottori mutant D7N (72), as well as the N-terminal truncated pyroglutamate A␤ may possess different ion binding properties and result in altered aggregation pathways.
It is particularly interesting that Zn 2ϩ induced preferentially ring-shaped, pore-like annular protofibrils without an apparent nucleation process and that Al 3ϩ induced similar oligomers. Recently, heterogeneous A␤ oligomers have been found to exist in immunological distinct structural states (67,73). These A␤ oligomers have been characterized as prefibrillar oligomers, fibrillar oligomers, and annular protofibrils. The annular protofibrils are considered to be assembled from the prefibrillar oligomers, where the conversion is accelerated in the presence of lipid vesicles and other artificial conditions. The annular protofibrils are less toxic and membrane-permeable in comparison with prefibrillar oligomers, but they share common structural epitopes with the bacterial toxin formed by ␤-barrel structures (67). In this study, the Zn 2ϩ -or Al 3ϩ -induced A␤ oligomers were morphologically similar to the annular protofibrils and also shared a common epitope recognized by A11 antibody. The extent to which the Zn 2ϩ -or Al 3ϩ -induced A␤ oligomers resemble those annular protofibrils or others remains uno oknown. The structural conversion, membrane permeability, and toxicity of those annular protofibrils need further structural and functional characterizations. However, our far-UV CD spectra of Zn 2ϩ -induced oligomers showed no enrichment of the ␤-conformation, which suggests that the oligomers are not fibrillar oligomers and should be different from the ␤-barrel toxins.
In physiological conditions, less than 0.5 M of extracellular Zn 2ϩ is present in brain; however, a high concentration of Zn 2ϩ was found in the glutamergic neurons during neuronal activity especially in cerebrocortex or amygdala (22,74). The presynapic Zn 2ϩ is concentrated by the zinc transporter protein, ZnT3, to achieve ϳ300 M in the cleft (22,75). The in vivo evidence shows that the ZnT3 knock-out transgenic amyloid precursor protein mice with a reduction of Zn 2ϩ in the hippocampus did not develop obvious amyloid plaques in comparison with those with normal ZnT3 level (34). In AD patients, a decrease of metallothionein 3 that regulates the uptake of synaptic Zn 2ϩ has been observed (35), and the release of synaptic Zn 2ϩ during activity is critical for A␤ oligomer synaptic targeting (36). These studies show that synaptic Zn 2ϩ and Zn 2ϩ homeostasis play crucial roles in AD pathology. According to our results, we suspect that the Zn 2ϩ -induced oligomer formation resulting in heterogeneous A␤ annular protofibrils may occur at presynapses during synaptic activities, thereby triggering oligomer formation and leading to concentrated A␤ oligomers at the synaptic clefts. This hypothesis may also contribute to the mechanism of rapid cognition restoration in the metal chelating therapy (53). In summary, our fundamental folding and aggregation studies facilitate the understanding of the clinically related metal ions in AD to A␤ and provide a rationalized mechanism for annular protofibril formation. The study may contribute to potential pathogenic implication in AD.