Truncated Amyloid-β(11–40/42) from Alzheimer Disease Binds Cu2+ with a Femtomolar Affinity and Influences Fiber Assembly*

Background: N-terminally truncated Aβ(11–40/42) typically constitutes 19% of plaque load in Alzheimer patients. Results: Cu2+ binds to Aβ(11–40) with a 34-fm dissociation constant and stabilizes very short highly amyloidogenic amyloid rods into longer Aβ fibers. Concussion: The very tight affinity explains the high levels of Cu2+ in amyloid plaques. Significance: Copper, tightly bound to Aβ(11–40/42), may influence disease pathology. Alzheimer disease coincides with the formation of extracellular amyloid plaques composed of the amyloid-β (Aβ) peptide. Aβ is typically 40 residues long (Aβ(1–40)) but can have variable C and N termini. Naturally occurring N-terminally truncated Aβ(11–40/42) is found in the cerebrospinal fluid and has a similar abundance to Aβ(1–42), constituting one-fifth of the plaque load. Based on its specific N-terminal sequence we hypothesized that truncated Aβ(11–40/42) would have an elevated affinity for Cu2+. Various spectroscopic techniques, complemented with transmission electron microscopy, were used to determine the properties of the Cu2+-Aβ(11–40/42) interaction and how Cu2+ influences amyloid fiber formation. We show that Cu2+-Aβ(11–40) forms a tetragonal complex with a 34 ± 5 fm dissociation constant at pH 7.4. This affinity is 3 orders of magnitude tighter than Cu2+ binding to Aβ(1–40/42) and more than an order of magnitude tighter than that of serum albumin, the extracellular Cu2+ transport protein. Furthermore, Aβ(11–40/42) forms fibers twice as fast as Aβ(1–40) with a very different morphology, forming bundles of very short amyloid rods. Substoichiometric Cu2+ drastically perturbs Aβ(11–40/42) assembly, stabilizing much longer fibers. The very tight fm affinity of Cu2+ for Aβ(11–40/42) explains the high levels of Cu2+ observed in Alzheimer disease plaques.

copper concentration is at least four times higher in AD brains, and Cu 2ϩ is bound directly to A␤ in plaques (25)(26)(27)(28). Changes in copper homeostasis exacerbate disease phenotype in rabbit and Drosophila models of AD (29 -31). Furthermore, sequestering copper with chelators such as clioquinol or its derivatives lowers A␤ aggregate load, maintains the health of AD mice, and reduces toxicity in models of AD (32)(33)(34). Cu 2ϩ is thought to bind to A␤ (1-40/42) with a 50 -100 pM dissociation constant (35)(36)(37), and there is a suggestion that A␤ (1-40/42) may bind Cu 2ϩ with a tighter affinity than this in an aggregated form (36). However, although the fluxes of weakly bound synaptic Cu 2ϩ are thought to reach levels of 15-200 M (38, 39), it is still contested as to whether a 50 pM affinity is sufficient to retain Cu 2ϩ ions in the brain parenchyma in the presence of other endogenous Cu 2ϩ chelators. We wondered whether the concentration of Cu 2ϩ in plaques might in part be due to Cu 2ϩ binding to A␤ (11-40/42) with an even tighter affinity.
Here we aim to characterize the amyloid-forming properties of A␤  and A␤  and their interaction with Cu 2ϩ . We show for the first time that A␤ (11-40/42) has heightened amyloidogenicity compared with A␤ (1-40/42) and forms short amyloid rod-shaped assemblies. Furthermore, we show Cu 2ϩ has an extremely tight, 34 Ϯ 5 fM dissociation constant for A␤  at pH 7.4 and has a marked influence on A␤  fiber formation and morphology. A␤ (11-40/42) constitutes 19% of the plaque load, and this work suggests that it is responsible for the high concentration of Cu 2ϩ found in the plaques of AD patients.
These peptides, found in the CSF and plaques, include a free N-terminal amino group and C-terminal carboxylate. Pentapeptide models of the Cu 2ϩ binding motif were C-terminally amidated to mimic the continuation of the peptide sequence in full-length A␤ (11-40/42) . The peptides were removed from the resin and deprotected before being purified by reverse-phase high performance liquid chromatography. The samples were characterized using mass spectrometry. Peptides with amidated C termini and free N-terminal amino groups studied included NH 2 -EVHHQ-CONH 2 and NH 2 -EVHAQ-CONH 2 . Pentapeptides were purchased from Generon Ltd. (Maidenhead, UK). Full-length A␤  , A␤  , and A␤  purchased from EZ Biolabs (Carmel, IN) and Cambridge Research Biochemicals (Billingham, UK).
A␤ Solubilization-All three A␤ peptides, A␤  , A␤  , and A␤  , were solubilized at a concentration of 100 M in water at pH 10.5-11 and then left at 4°C for 48 -72 h. In our hands we found that this was the most effective solubilization method, as others have reported (47,48). It is clear that solubilized preparations of A␤  and A␤  are largely seedfree at pH 7.4, as solubilized preparations typically have no thioflavin T (ThT) fluorescence at time 0 and exhibit a distinctive lag phase of Ͼ20 h for A␤  and 60 h for A␤ . Freshly solubilized A␤  and A␤  peptides had no detectable assemblies under TEM. For A␤  , the concentration was determined using tyrosine absorption at 280-nm ⑀ 280 ϭ 1280 cm Ϫ1 M Ϫ1 . A␤  and A␤  lack a tyrosine residue so the peptide absorbance at 214-nm ⑀ 214 ϭ 923 cm Ϫ1 M Ϫ1 per amide was used (49). All lyophilized peptides typically contained 10 -20% moisture by weight. Concentrations for A␤ (11-40/42) were supported by the 1:1 saturation of the Cu 2ϩ binding site, indicating that a very clear 1:1 stoichiometry had been established for freshly solubilized A␤ (11-40/42) .
Titrations-All chemicals were purchased from Sigma at the highest purity available, and ultra high quality water was used throughout (resistivity of 10 Ϫ18 ohm Ϫ1 cm Ϫ1 ). Small aliquots of fresh aqueous stock solutions, typically 10 mM, were used to add metal ions (CuCl 2 ⅐2H 2 O, NiCl 2 ⅐6H 2 O) and competitive ligands, glycine (500 mM) and L-histidine (50 mM), stocks for titrations. Titrations were conducted at pH 7.4 with adjustments made using 20 -100 mM solutions of NaOH and HCl. pH measurements were made before and after spectroscopic measurements.
Affinity Measurements-The conditional affinities at pH 7.4 were determined using glycine and L-histidine as competing ligands. The range of competitor concentrations (glycine and histidine) with Cu 2ϩ -A␤  or Cu 2ϩ -A␤ (11)(12)(13)(14)(15) (0.8:1.0) complexes were made up in advance and left for 12 h at pH 7.4 to ensure equilibrium had been reached. Repeat CD spectra were obtained over a few hours to confirm equilibrium had FIGURE 1. ␤and ␤-secretase processing of APP. The ␤Ј activity of the ␤-site amyloid precursor protein cleaving enzyme 1 (BACE-1) N-terminally truncates A␤ before Glu 11 . ␥-Secretase cleaves the C terminus most commonly at position 40 or 42, within the lipid bilayer. The ␣-secretase produces P3, which starts at Lys-17. The percentages of typical A␤ plaque load are also indicated; 19% for A␤ (11-40/42) . been reached. The concentration of competitor required for equal molar equivalents of Cu 2ϩ to be bound to both the A␤ peptide and competitor was used to determine the affinity of Cu 2ϩ for the A␤ peptide using Equation 1 (35,37,50). Cu 2ϩ free is the concentration of Cu 2ϩ not bound to either competitor ligand or peptide.
Fiber Growth Assay-The kinetics of amyloid fiber formation were monitored using binding of ThT to amyloid fibers, which induces ThT to fluoresce at 487 nm; this signal is proportional to the amount of amyloid fibrils present (52). BMG-Galaxy and Omega FLUOstar fluorescence 96-well plate readers were used for the measuring of ThT fluorescence. Fluorescence readings were typically taken every 30 min after 30 s of gentle agitation. Fluorescence excitation and emission detection were obtained using filters at 440 nm and 490 nm, respectively. Fiber growth kinetics are sensitive to a number of factors, including pH, concentration of A␤, agitation, ionic strength, and temperature; consequently, measures were taken to reduce variance in these parameters. All A␤ peptides were studied at 10 M, and all fiber growth experiments were incubated at 30°C in 30 mM HEPES buffer (due to its low affinity for Cu 2ϩ ions) and 160 mM NaCl. The pH, a critical factor in rate of fiber growth, was adjusted to pH 7.4 if necessary with small additions of 20 -100 mM NaOH and HCl. Variation between samples was measured to be 0.05 pH units or less. 10 M ThT was present (1:1 with A␤), which is optimal for tracking A␤ amyloid growth (52).
Growth Curve Analysis-Conversion of essentially monomeric A␤ to amyloid fibers follows a characteristic growth curve consisting of an initial lag-phase (nucleation) before a growth phase (elongation). A growth curve can be fitted to the data to obtain a number of empirical parameters using the following equation (53).
y represents fluorescent intensity, and x represents time. Initial fluorescence intensity is represented by v i , v f represents the final fluorescence intensity, and x 0 is the time at which halfmaximal fluorescence is reached (t 50 ). The conditional fiber growth rate (k app ) is obtained by 1/ and the lag-time (t lag ) by x 0 Ϫ 2. Data were processed using KaleidaGraph 4.0. Typically six or more kinetic traces were used to extract kinetic parameters. Circular Dichroism (CD)-CD spectra were recorded at 25°C on an Applied Photophysics Chirascan instrument between 250 and 700 nm with sampling points every 2 nm using a 1-cm path length. Three repeat scans were made, and baseline spectra were subtracted from each spectra followed by smoothing using a 5-nm window if required. Data were processed using Applied Photophysics Chirascan Viewer, Microsoft Excel, and KaleidaGraph spreadsheet/graph package. CD ellipticity measurements (, millidegrees) were converted to ⌬⑀ (molar CD, M Ϫ1 cm Ϫ1 ) using the relationship ⌬⑀ ϭ /(33000.cl), where c is the molar concentration, and l is the path length. Note, molar ellipticity [] (degree cm Ϫ1 dmol Ϫ1 ) ϭ ⌬⑀ 3300.
TEM-Aliquots of A␤ samples from the fiber growth assays were added to glow-discharged carbon-coated 300-mesh copper grids (Structure Probe, Inc. West Chester, PA) using the droplet method, with water washes before and after the addition of stain. Phosphotungstic acid (2% w/v), adjusted to pH 7.4, was used to negatively stain the assemblies. Phosphotungstic acid was prepared and incubated at 37°C for 24 -48 h before filtering immediately before use. Selected images are representative of multiple images that were taken over a number of fields and grids. Images were recorded using a JEOL JEM-1230 electron microscope operated at 80 kV and the Olympus iTEM software package.
Electron Paramagnetic Resonance (EPR)-EPR was carried out on a Bruker Elexsys E580 spectrometer at an X-band (9.38 GHz) microwave frequency at 0.5 milliwatt. Single scans were made in EPR quartz tubes (outside diameter 4 mm, inside diameter 3 mm) with a 2000 G sweep width centered at 3000-G and a 10-G modulation amplitude. 50 mM 60% HEPES, 40% phosphate buffer was used at pH 7.4 for temperature-independent buffering at 10 K temperatures. Hyperfine splitting, A II (millikaisers, mK) ϭ 0.046686 ϫ g ϫ ⌬H, where g ϭ 2.0023, and ⌬H is the A II splitting, measured in Gauss.
1 H NMR-1 H NMR spectra were recorded on a Bruker AV600 spectrometer using 500 M EVHHQ peptide in 100% D 2 O, 30 mM phosphate buffer at pH* 7.4 (pH* denotes the pH reading taken in 100% deuterium (D 2 O)). CuCl 2 aliquots were added from 10 mM. Stocks of CuCl 2 ⅐2H 2 O were in 100% D 2 O. Data were processed and analyzed using TopSpin 2.1 (Bruker) software.

Results
A␤  and A␤  Rapidly Forms Amyloid Rods-Although N-terminally truncated A␤  and A␤  constitute 20% of AD plaque load, very little is known about their ability to form amyloid fibers. We, therefore, monitored the kinetics of A␤  and A␤  fiber formation using the amyloid-specific fluorescent dye, ThT. After solubilization at high pH, A␤  was diluted to 10 M at pH 7.4, and ThT fluorescence was monitored over time in a 96-well plate reader. The kinetics of A␤  fiber formation were compared with A␤  . Like A␤  , a sigmoidal fiber growth curve was consistently observed for A␤  with a clear nucleation and elongation phase, shown in Fig. 2A. A␤  forms fibers significantly faster than A␤ , with an accelerated nucleation and elongation rate. Under the same conditions A␤  has a nucleation time of 27 Ϯ 0.3 h compared with 65 Ϯ 1 h for A␤  . The elongation rates for A␤  and A␤  were 0.44 Ϯ 0.02 and 0.23 Ϯ 0.04 h Ϫ1 , respectively. Total maximal ThT fluorescence for the two A␤ forms were similar at 10 M, suggesting both form comparable amounts of amyloid, detectable by ThT.
Next we wanted to investigate the types of A␤ assemblies generated. TEM of A␤  , shown in Fig. 2B, consistently revealed stacking of straight, very short rod-shaped fibers. Only short, stacked rods were observed across multiple peptide batches, solubilizations, and TEM grids. The rods had a similar thickness (ca. 10 nm) to A␤  fibers ( Fig. 2C) but were much shorter, with the majority being under 50 nm in length. These amyloid rods did not extend beyond 500 nm, as shown in Fig.  2D. Furthermore, individual fiber rods tend to self-associate into clumps much more readily than that of the typically long, twisted A␤  fibers, shown in Fig. 2c, formed on the same well-plate under the same conditions. A␤  was also solubilized at high pH with the method used for A␤  . The ThT fluorescence of A␤  produced a sigmoidal growth curve with a nucleation and elongation phase. Under the same conditions, both A␤  and A␤  have quite a similar nucleation phase with lag times of 27 Ϯ 0.3 h and 25 Ϯ 1 h, respectively, shown in Fig.  2a. A␤  has some fluorescence signal directly after solubilization, suggesting some amyloid assemblies persist even after solubilization.
TEM images revealed some different morphological features for A␤  compared with A␤  . A␤  contained short stacked rods just like A␤   (Fig. 2E), but there were also longer twisted fibers ( Fig. 2F) often with ragged oligomeric assemblies along the length of the fiber. Also present are some highly stained disordered amorphous aggregates, which are in greater abundance in A␤  than A␤  images.
We were interested in why A␤  formed much shorter fiber rods. We wondered if A␤  rods were caused by fiber fragmentation so we generated A␤  fibers in the absence of the mild agitation used in the previous experiment. Quiescently generated fibers showed a clear change in the fiber length of A␤  . Although numerous short amyloid-rods were present, much longer fibers (which were not observed under mild agitation conditions) were also observed (Fig. 3a). This suggests that A␤  forms exclusively short amyloid rods because it readily fragments. The differences in quiescently generated A␤  fibers were less marked, but the short rods were less abundant, with a greater relative abundance of longer fibers (Fig. 3b).
Cu 2ϩ Tetragonal Coordination to A␤  -There is a well established link between Cu 2ϩ homeostasis and AD; we, therefore, wanted to investigate the Cu 2ϩ binding properties of A␤ (11-40/42) . For this we have used visible CD and EPR spec-   and A␤  form amyloid rods. A, ThT fluorescence follows the fiber growth of A␤  (blue), A␤  (green), and A␤  (red). Shown are representative TEM images of A␤  (B), A␤  (C), and A␤  (E and F) negatively stained with phosphotungstenic acid; the scale bar is 200 nm. Fiber lengths were measured for A␤  (blue) and A␤  (red), and their percentage frequency is shown in d. N-terminally truncated A␤ formed fibers more rapidly than full-length A␤ and produce short stacked amyloid rods, mild intermittent agitation for all conditions, 10 M peptide, 10 M ThT, 30 mM HEPES, and 160 mM NaCl at pH 7.4.
troscopy. When bound to proteins and peptides Cu 2ϩ can produce very characteristic visible CD bands (54,55). Cu 2ϩ -derived visible CD bands arise from amide-main chain coordination being in close proximity to the chiral center, fixed in a chelate ring (56,57). Fig. 4A showed an overlay of the visible CD spectra of the N-terminal residues for A␤  and a more soluble N-terminal pentapeptide A␤ (11)(12)(13)(14)(15) (NH 2 -EVHHQ-Am) with 1:1 mol eq of Cu 2ϩ . The spectra are almost identical, suggesting that only the first few N-terminal residues of A␤  are involved in coordinating Cu 2ϩ ions.
In light of the possible similarity between Cu 2ϩ binding at the N terminus of HSA and A␤  (in particular, both contain the NH 2 -(Xaa) 6 His Cu 2ϩ binding motif), we directly compared the visible CD spectra of Cu 2ϩ -loaded A␤  with that of HSA and two albumin peptide models: NH 2 -Asp-Ala-Hisamide and NH 2 -Ala-Ala-His-amide. It is clear from Fig. 4B that the visible CD spectra are very similar. There are some minor differences in the relative intensity of the positive and negative CD bands, but in comparison to the wavelength shifts seen between HSA and its N-terminal peptide models, these are not significant enough to suggest a different binding conformation (58).
The stoichiometry of the Cu 2ϩ -A␤  interaction was investigated though a CuCl 2 titration monitored by visible CD with A␤  and A␤ (11)(12)(13)(14)(15) , shown in Fig. 4C. The CD band intensity increased with the addition of Cu 2ϩ up to 1 mol eq, indicating a 1:1 stoichiometry. A single set of CD bands was observed throughout the titration, suggesting a single complex was formed even at low, substoichiometric amounts of Cu 2ϩ . The Cu 2ϩ complex was stable over a large physiological pH range, with a mid-point for the complex of 4.7 at the transition between the CD active Cu 2ϩ complex and no CD signal. Fig. 4D showed the pH dependence of the Cu 2ϩ -A␤ (11)(12)(13)(14)(15) complex. There was no change in the wavelength position of the CD bands, indicating that a single CD active complex was formed between pH 5.5 and 10. It is clear the same Cu 2ϩ complex formed over a large range of pH values.
ited Cu 2ϩ type-II spectra typical of a square-planar/tetragonal coordination geometry, shown in Fig. 5A. Furthermore the g II of 2.17 and hyperfine splitting, A II , of 19.3 mK is most indicative of four nitrogen coordinating ligands (59). The Cu 2ϩ -A␤ (11)(12)(13)(14)(15) EPR spectra bare a close resemblance to the EPR spectra of the albumin N-terminal Cu 2ϩ binding site, also shown in Fig. 5A. The g II and A II values are very similar, further supporting that the Cu 2ϩ -A␤ (11)(12)(13)(14)(15) complex has a square-planar, four-nitrogen coordination similar to that found for albumin, with the N-terminal amino group, the next two amide main-chain nitrogens, and the imidazole ring nitrogen coordinating, as shown in Fig. 5B.
We were interested in investigating if Cu 2ϩ would bind to the amyloid fibril rods of A␤  in a similar manner to the monomer . Using visible CD we show that Cu 2ϩ was able to bind to preformed A␤  fiber rods. The visible CD spectra bears a close resemblance to the Cu 2ϩ -A␤  monomer complex with comparable wavelengths and intensity of visible CD bands, as shown in Fig. 6. Fibrillar Cu 2ϩ -A␤  exhibited 0.7:1 binding stoichiometry close to 1:1. This suggests the Cu 2ϩ binding complex within the first three residues at the N terminus is accommodated within the amyloid fiber structure. We note from this binding curve a 0.5:1 stoichiometry, which although unlikely, cannot be ruled out in fibers.
Femtomolar Cu 2ϩ Conditional Dissociation Constant for A␤  -The affinity of Cu 2ϩ binding to A␤  and its more soluble peptide model A␤ (11)(12)(13)(14)(15) was determined using a number of competitive ligands with known affinities for Cu 2ϩ . The tight affinity of Cu 2ϩ for A␤  requires a highly soluble competitor so glycine and histidine were chosen for their relatively high solubility. Glycine has a micromolar affinity for Cu 2ϩ and forms a bidentate complex via its amino and carboxylate groups, whereas histidine has a nanomolar affinity and can coordinate Cu 2ϩ , forming a Cu 2ϩ (His) 2 bidentate complex at physiological pH (60).
The concentration of competing ligand required to remove half of the Cu 2ϩ from A␤ (11)(12)(13)(14)(15) could be used to determine the affinity using Equation 1; see "Experimental Procedures." The molar ellipticity from the visible CD spectra bands at 276, 316, 480, and 558 nm was used to gain an insight into the errors in fitting the binding curves. In addition, the titrations were repeated for each competitive ligand. Table 1 shows affinities determined with histidine and glycine. Using glycine as a competitor, a K d of 2.5 ϫ 10 Ϫ14 M was determined at pH 7.4, whereas for histidine a K d of 4.2 ϫ 10 Ϫ14 M was calculated. An analysis of variance showed there was no significant difference in the values determined using histidine and glycine as competitors with a p value of 0.20 across the 4 titrations. It is clear that the conditional dissociation constant at pH 7.4 of Cu 2ϩ for A␤ (11)(12)(13)(14)(15) is in the femtomolar range with a mean K d of 34 Ϯ 5 fM determined. The agreement in affinity determined using two different competing ligands indicated the measurements were not complicated by the formation of a ternary complex between Cu 2ϩ , A␤ (11)(12)(13)(14)(15) , and competitor.
A similar glycine competition titration of the full-length monomeric Cu 2ϩ -A␤  complex was performed, shown in Fig.  7C. A similar K d of 10 Ϯ 3 fM was measured. This data were less reliable, as A␤  was less soluble, and there were some light scatter in the data that affects the visible CD signal; however, it was clear A␤  bound to Cu 2ϩ with an affinity of 34 fM or tighter.
Evidence for Square-pyramidal His-14 Coordination of Cu 2ϩ -The affinities determined for A␤  in Table 1 are tighter than those reported for human serum albumin, and we considered the possibility that the higher affinity might be due to the presence of a second histidine residue, His-14, four residues from the N terminus. To investigate this we generated a peptide analogue with an alanine substitution at the His-14 site. Using competing ligands, we determined the affinity of Cu 2ϩ for A␤ (11)(12)(13)(14)(15) H14A. In this way we have determined if His-14 has an influence on the Cu 2ϩ affinity. The dissociation constant at pH 7.4, determined using glycine or histidine as competing ligands, was 136 Ϯ 23 fM. This value was slightly weaker than the complex with His-14 present, with a p value of Ͻ0.05 showing a significant difference between His-14 and alanine-substituted A␤ (11)(12)(13)(14)(15) affinities. Based on the K d data, it does appear His-14 has an influence on the affinity of Cu 2ϩ for A␤  .
The Cu 2ϩ visible CD and UV-visible absorption spectra of A␤ (11)(12)(13)(14)(15) are very similar compared with the H14A analogue, with a high degree of similarity between the visible CD band wavelengths and intensities. In addition, the visible absorption spectra for Cu 2ϩ -A␤ (11)(12)(13)(14)(15) and A␤ (11)(12)(13)(14)(15) H14A are very similar. A max at 525 nm and a molar extinction coefficient ⑀ 525 of 110 M Ϫ1 cm Ϫ1 for both peptides are indicative of other NH 2 -XXH Cu 2ϩ complexes (45). This further supports the forma-FIGURE 6. Cu 2؉ binds to A␤  fibers. A comparison of visible CD spectra of Cu 2ϩ loaded monomer (red) and fibers (blue) of A␤  . The inset shows A␤  fibers loading Cu 2ϩ with approximate 1:1 stoichiometry (0.7: 1). Preformed A␤  fibers (70 M) were generated before Cu 2ϩ titration, pH 7.4. A␤  fibers can accommodate Cu 2ϩ binding in the same manner as the monomeric form, with similar CD bands and intensities.
tion of a characteristic NH 2 -XXH Cu 2ϩ complex similar to that of the N-terminal Cu 2ϩ binding site of albumin.
We wanted to use 1 H NMR to help further resolve this question. The addition of paramagnetic Cu 2ϩ causes a loss of A␤ (11)(12)(13)(14)(15) 1 H NMR signal intensity with increasing additions of Cu 2ϩ . Cu 2ϩ -bound 1 H NMR signals were not observed due to the paramagnetic properties of Cu 2ϩ , which cause significant linebroadening of 1 H NMR signals. There was no difference in the extent to which signal was lost between His-13 and His-14 resonances as Cu 2ϩ was added, which is consistent with a role for His-14 in axial coordination. However, this could be due to the proximity of His-14 ring protons to Cu 2ϩ rather than direct coordination, as all the side-chain resonances are extensively broadened.
Cu 2ϩ and A␤  Fiber Formation-We have shown A␤  has an appreciable femtomolar affinity for Cu 2ϩ ; therefore, we wanted to investigate how the presence of Cu 2ϩ ions might affect the kinetics and morphology of A␤  fiber formation. ThT fluorescence indicates that the presence of 0.1 mol eq of Cu 2ϩ has little effects on fiber kinetics, whereas 0.2 mol eq causes a reduction in the total ThT signal by ϳ40%; see Fig. 8A. A similar ThT signal reduction was observed with 0.4 mol eq of Cu 2ϩ , but there was a prominent reduction in elongation rate, with k app slowing from 0.54 Ϯ 0.02 h Ϫ1 for apo to 0.21 Ϯ 0.04 h Ϫ1 . For 1 and 10 mol eq of Cu 2ϩ , no ThT signal was observed over Ͼ100 h. Interestingly, not only do small substoichiometric amounts of Cu 2ϩ perturb the kinetics of A␤  fiber formation and the total amount of ThT bound to amyloid, but Cu 2ϩ also affects the morphology of the fibers generated. In the absence of Cu 2ϩ , A␤  exclusively generates short amyloid rods (mostly 50 -200 nm in length) under mildly agitating conditions, as shown in Fig. 2. However, with fibers generated under the same mild agitation, the TEM images show that short rods only partially remain even in the presence of small substoichiometric concentrations of 0.2 eq of Cu 2ϩ . However, there are also similar amounts of much longer, twisted fibers with a morphology more reminiscent of A␤  , shown in Fig. 8B. At just 0.4 eq of Cu 2ϩ , EM grids contain almost exclusively long fibrillar assemblies. Counterintuitive, although at 0.4 mol eq of Cu 2ϩ the ThT signal is attenuated, the EM images show numerous and wide-spread long amyloid fibers (Fig. 8C), but there was an increased amount of amorphous aggregates compared with lower Cu 2ϩ equivalents. Beyond 1 mol eq of Cu 2ϩ the TEM images were dominated by amorphous aggregates with dense irregular staining of the aggregates (not shown), which do not cause ThT fluorescence.
We were also interested in discovering if the long fibers generated in the presence of Cu 2ϩ revert to short rods in the presence of a strong Cu 2ϩ chelator; for this experiment we added EDTA, which has a K a of 10 18 for Cu 2ϩ . The addition of EDTA to A␤  fibers, generated in the presence of 0.4 mol eq of Cu 2ϩ , caused the ThT fluorescence signal to increase to a similar intensity to that observed for apo A␤  , as shown in Fig.  9A; thus, any reduction in ThT bound fibers is rapidly reversible. Interestingly TEM images of these A␤  fibers grown with Cu 2ϩ revert to very short stacks of amyloid rods like those seen for fibers of A␤  only. This observation suggests that Cu 2ϩ has a stabilizing effect on A␤  fibers, preventing fragmentation, but once Cu 2ϩ was removed, all the fibers rapidly fragmented into amyloid rods, as shown in Fig. 9B.
Next we wanted to see if Cu 2ϩ addition affects the morphology of preformed A␤  amyloid rods as well. We know from Fig. 6 that Cu 2ϩ can bind to preformed amyloid fibers. The addition of Cu 2ϩ to preformed amyloid rods of A␤  caused a 32% reduction in ThT fluorescence. This reduced the intensity of fluorescence to a similar level to the maximum fluorescence observed for A␤  fibers grown in the presence of 0.4 mol eq of Cu 2ϩ . TEM images, shown in Fig. 9C, demonstrate that the addition of Cu 2ϩ to the preformed, fragmented amyloid rods had no discernible effect over a 70-h period.

Peptide
Glycine Histidine Mean fM A␤ (11)(12)(13)(14)(15) 27 Ϯ 3 and 24 Ϯ 3 50Ϯ 16 and 34 Ϯ 2 34Ϯ 5 A␤  10 Ϯ 3 1 0 Ϯ 3 fM A␤ (11)(12)(13)(14)(15)  A␤  only protofibrils and oligomers were observed. We, therefore, wanted to investigate the effect of Cu 2ϩ on A␤  fiber growth to assess how the two C-terminal amino acids influence Cu 2ϩ -dependent kinetics. As with A␤  , increas-ing the mol eq of Cu 2ϩ causes the maximum ThT signal of each A␤  fiber growth curve to decrease. The ThT signal reduction caused by 0.4 mol eq of Cu 2ϩ is similar to A␤  , with a reduction for A␤  of ca. 50%. The effect of Cu 2ϩ on A␤  FIGURE 8. Influence of Cu 2؉ on A␤  and A␤  fiber growth. ThT Fluorescence of A␤  and A␤  fiber growth with different molar equivalencies of Cu 2ϩ , A and D, respectively. TEM images of fibers produced with the following Cu 2ϩ equivalencies: A␤  0.1 (B) and 0.4 (C); A␤  0.1 (E) and 0.4 (F). All scale bars are 50 nm. AFU, arbitrary fluorescence units. . Reversibility of Cu 2؉ influence on A␤  fibers. A, shown is the effect of 0.4 mol eq Cu 2ϩ on preformed A␤  fibers and removal of Cu 2ϩ from Cu 2ϩ -A␤  fibers. Shown is ThT fluorescence of A␤  fiber growth with the subsequent addition of 0.4 mol eq of Cu 2ϩ at 235 h and fiber growth of A␤  in the presence of Cu 2ϩ with the later addition of 0.4 mol eq of EDTA. ThT fluorescence kinetics traces are an average from n ϭ 6; error bars are S.E. Shown are TEM images of the fibers produced with Cu 2ϩ added to preformed fibers (B) and Cu 2ϩ -A␤  fibers with subsequent EDTA addition (C). Scale bars are 200 nm. AFU, arbitrary fluorescence units. elongation rate was similar to A␤  with a slower rate with increasing Cu 2ϩ levels.
The representative TEM images of A␤  nucleated with Cu 2ϩ , shown in Fig. 8, E and F, complemented the decrease in ThT fluorescence observed, as A␤  was grown with higher equivalencies of Cu 2ϩ . There was a general decrease in short amyloid rod stacks, while longer fibers were more abundant as more Cu 2ϩ was added. There were also shorter, "curvy" assemblies with a morphology often observed for protofibrillar assemblies at 0.4 mol eq of Cu 2ϩ . The promotion of A␤  protofibril generation by Cu 2ϩ is similar to the effect observed for Cu 2ϩ with A␤  , but the effect was by no means as universal as it is for A␤   (24).

Discussion
A␤ (11-40/42) Rapidly Forms Amyloid Rods-Appreciable levels of A␤ (11-40/42) are present in the CSF, produced by the action of BACE at the ␤Ј cleavage site of APP (10,11). Furthermore, mean levels of A␤ within plaques from sporadic AD patients have been shown to contain 19% A␤  , comparable to A␤  levels (12,13). Despite this, perhaps because of its poor solubility (18), very few in vitro studies of A␤ (11-40/42) have been reported, with studies focusing almost exclusively on A␤ (1-40/42) . This is the first study describing the kinetics of A␤ (11-40/42) fiber formation and its interaction with Cu 2ϩ ions.
Under the same conditions (10 M A␤, pH 7.4, mild agitation) both A␤  and A␤  form ThT binding amyloid assemblies more than twice as fast as A␤  . Thus the rapid formation of amyloid rods of A␤ (11-40/42) that we observed may be important in plaque deposition. The amyloidogenicity of A␤ sequences is often thought to correlate with disease progression. A␤  forms fibers much faster than A␤  and is also more toxic in vivo (61)(62)(63)(64). The heightened rate of fiber nucleation and elongation observed for A␤ (11-40/42) may be caused by the loss of negatively charged side chains, as it is known that the pI and charge of A␤ have a profound influence on its rate of fiber formation (22,65). The absence of a number of charged residues at the N terminus increases the theoretical pI from 5.1 for A␤ (1-40/42) to 6.0 for A␤ (11-40/42) . Consequently, at physiological pH 7.4, A␤  is more neutrally charged and will much more readily self-associate into nucleating oligomers and elongate more rapidly into fibers. A␤  fibers generated quiescently form much longer fibers indicating A␤  fibers fragment more easily than A␤  . It is known that secondary nucleation events can dominate the kinetics of fiber formation and influence cytotoxicity (66,67). The rapid fiber formation for A␤  might be due to its propensity to fragment and cause secondary nucleation.
We were interested by the markedly shorter rod-like fibers observed for apoA␤  and A␤  compared with A␤  . With the exception of the A␤ Osaka mutant (A␤⌬22) (68), the first 10 residues of A␤ (1-40/42) have been shown not to typically be directly involved in the core structure of most amyloid fibers generated in vitro and are not thought to form any stable hydrogen-bonded structure (69 -71); despite this, our data show these residues have a profound influence on the appearance (length) of the fibers. It is possible that the N-terminal residues (1-10) have a stabiliz-ing influence on the amyloid fibers as they form. A␤  also forms largely short amyloid rods under mildly agitating conditions; although longer fibers are present, the two additional C-terminal residues could possibly stabilize the intermolecular forces along the fiber.
Cu 2ϩ and A␤ (11-40/42) -The presence of the histidine three residues from the N terminus facilitates the formation of a high affinity Cu 2ϩ complex. A␤  and the model peptide A␤ (11)(12)(13)(14)(15) have a 34 Ϯ 5 fM conditional dissociation constant for Cu 2ϩ at pH 7.4. This indicates A␤ (11-40/42) binds to Cu 2ϩ with a 1000ϫ tighter affinity compared with A␤ (1-40/42) (35). It is clear that A␤ (11-40/42) will preferentially bind any available extracellular Cu 2ϩ ions. Indeed A␤  has an affinity more than an order of magnitude tighter than the extracellular Cu 2ϩ transport protein HSA. This has ramifications for the ability of A␤ (11-40/42) to compete for Cu 2ϩ ions at the synapse. Cu 2ϩ is released at the synapse during neuronal depolarization and is thought to reach levels between 15 and 200 M (38). CSF levels of Cu 2ϩ are lower, at 500 nM, (72), but the femtomolar affinity of A␤ (11-40/42) will readily bind Cu 2ϩ even at these lower concentrations. The total soluble A␤ concentration in the CSF is 0.1-0.5 nM (11), so with such a high affinity for Cu 2ϩ much of the A␤ (11-40/42) is likely to coordinated Cu 2ϩ ions. It is notable that the most abundant protein in the CSF, the Cu 2ϩ transport protein serum albumin, reaches levels of 3 M but has an affinity for Cu 2ϩ more than an order of magnitude weaker than A␤ (11-40/42) and will, therefore, only partially compete for Cu 2ϩ ions bound to A␤ (11-40/42) . A␤ (11-40/42) can constitute up to 19% of the plaque load (12), and its presence can explain the very high levels of Cu 2ϩ found in the plaques of AD patients (25)(26)(27)(28).
The visible absorbance, visible CD, EPR spectra, pH dependence, stoichiometry, and NMR data are all initially consistent with the formation of a 4N, N-terminal complex shown in Fig.  5B. We considered the possibility of the involvement of the imidazole nitrogen of His-14 coordinating axially, perpendicular to the 4N plane. There is little difference in the absorbance spectra or visible CD between the A␤ (11)(12)(13)(14)(15) model peptide and A␤ (11)(12)(13)(14)(15) H14A, which might suggest His-14 is not involved in coordination; however, axial coordination can have minimal influence on absorbance spectra for Cu 2ϩ square-pyramidal complexes. There is a significant loss of affinity with the removal of the second histidine, which suggests that His-14 has an influence on coordination, forming a square pyramidal complex with axial coordination from the imidazole of His-14, making the affinity of A␤ (11-40/42) considerably tighter than that observed for serum albumin (HSA) at 1.0 pM (46).
The visible CD spectra generated for Cu 2ϩ -A␤  monomers and fibers are almost identical, indicating that Cu 2ϩ binds to fibers with a similar coordination to that of the monomer. Visible CD spectra are extremely sensitive to coordination; the intensity and even sign of the CD bands will be greatly influenced by even small changes in geometry (57,58). Cu 2ϩ crosslinking between two adjacent A␤ molecules is not supported by the data. In particular, the loss of main-chain amide coordination to be replaced by side-chain coordination from a histidine imidazole of a second A␤ molecule is likely to cause a profound change in the appearance of the CD bands, which is not observed. The ability of amyloid fibers to accommodate Cu 2ϩ binding has also been observed for A␤  and A␤   (35,(73)(74)(75)(76). Interestingly, Cu 2ϩ can diffuse into and load on to all A␤ molecules within the fibers and facilitate the Cu 2ϩ coordination with close to 1:1 stoichiometry.
It is not clear why the presence of Cu 2ϩ would promote markedly longer fibers compared with the much shorter, rodlike fibers observed for apoA␤  , generated under the same conditions of mild agitation. The binding of Cu 2ϩ ions appear to stabilize longer fibers that would otherwise fragment into shorter rods. This is supported by the effect of EDTA addition to Cu 2ϩ -A␤  fibers that results in rapid fragmentation of A␤  into shorter amyloid rods. The molecular packing of A␤  fibers in the presence of Cu 2ϩ seems to produce a greater fiber stability against shearing forces that cause fragmentation (66). We suggest that at pH 7.4 the addition of Cu 2ϩ will add a positive charge to A␤  and with a pI of 6.0 A␤  may be closer to charge neutrality, which might stabilize fibers. Potentially the Cu 2ϩ complex could cross-link between the His-13 and His-14 residues of two adjacent A␤ molecules within a fiber, stabilizing the extension of the fiber resulting in longer structures; however, there is no evidence for a change in coordination geometry for A␤  monomer or fibers, so this seems unlikely.
The marked influence of the two C-terminal residues (Ala-41-Val-42) on A␤ assembly and morphology in the presence of Cu 2ϩ has implications for cytotoxicity and membrane disruption. Cu 2ϩ -loaded A␤  generates some curvy, short assemblies reminiscent of the protofibril and oligomer structures observed exclusively when substoichiometric Cu 2ϩ is added to A␤  fibers (24).
In conclusion, the N-terminally truncated form of A␤ constitutes relatively high levels in the CSF and plaques of Alzheimer patients (11)(12)(13). Here we show A␤ (11-40/42) has marked amyloidogenicity and a femtomolar affinity for Cu 2ϩ . The affinity of Cu 2ϩ for A␤ has been a source of much attention, with much debate surrounding the ability of A␤ (1-40/42) to retain Cu 2ϩ in vivo with a 50 pM affinity (35). The 34 fM conditional affinity of A␤ (11-40/42) for Cu 2ϩ is considerably tighter than extracellular competitors such as serum albumin. We suggest the high abundance (19%) of A␤ (11-40/42) in plaques (12,13) must contribute to the high levels of Cu 2ϩ reported in plaques (25,27,28). Furthermore a second N-terminally truncated form of A␤, residues 4 -40/ 42, is also found in plaques and typically makes up 6% of plaque load (12). This peptide also contains a histidine at the third position from the N terminus and has recently been shown to bind Cu 2ϩ with a subpicomolar affinity (77). A further N-terminally truncated form, A␤(3-40/42), is also found within plaques and is thought to constitute 2% of plaque load (12). Its Cu 2ϩ binding properties have been carefully described (78). This complex does not have a histidine at the third position and forms a Cu 2ϩ complex more similar to A␤(1-40/42). Cu 2ϩ within the plaques can convert A␤  fibers into protofibrils/oligomers, which suggests a mechanism to generate more toxic A␤ species (24). Furthermore, the concentration of redox active Cu 2ϩ within plaques can explain the high levels of methionine sulfoxide as well as dityrosine formation observed in A␤ plaques (27,79). The femtomolar affinity of A␤ (11-40/42) for Cu 2ϩ has implications for chelation thera-pies targeting extracellular Cu 2ϩ and the mode of action of clioquinol and its derivatives (32,34).
Author Contributions-J. D. B. performed the experimental work, interpretation of data, and preparation of the manuscript. J. H. V. supervised planning and research, data interpretation, and preparation of the manuscript.