Alzheimer's Aβ40 Studied by NMR at Low pH Reveals That Sodium 4,4-Dimethyl-4-silapentane-1-sulfonate (DSS) Binds and Promotes β-Ball Oligomerization*

The Alzheimer's Aβ40 peptide forms soluble oligomers that are extremely potent neurotoxins and strongly impede synapses function. In this study the formation and structure of the large, soluble, neurotoxic Aβ40 oligomer called “β-ball” were characterized by two-dimensional NMR, circular dichroism, fluorescence spectroscopy, hydrogen exchange, and equilibrium sedimentation. In acidic aqueous solution, half the Aβ40 molecules are in the β-ball state; the remainder are monomeric. The equilibrium between the two states is slow as judged by NMR linewidths and is stable for months. The kinetics of β-ball formation from monomer are biphasic with τ1 = 7 min and τ2 = 80 min with no transient helix formation. Monomeric Aβ40 is essentially devoid of stable secondary structure, although the central, Leu17–Ala21, and C-terminal, Gly29–Val40, hydrophobic regions show propensity toward adopting extended structure, and residues 22–25 tended to form a turn. We found that sodium 4,4-dimethyl-4-silapentane-1-sulfonate (DSS) binds to the central hydrophobic region of monomeric Aβ40. DSS binds β-balls more strongly and caused them to double in size. Plausible micelle-like models for the β-ball structure with and without bound DSS are presented.

produced in vivo when the large, membrane-bound amyloid precursor protein is cleaved by ␤and ␥-secretase complexes (2,3). Harmless, monomeric A␤ can associate in vitro to form a series of long lived soluble oligomers before adopting the distinct fibril conformation present in amyloid plaques (4). The last years have witnessed a paradigm shift away from amyloid fibrils and toward soluble oligomers of A␤ as the conformation responsible for the loss of synapse function occurring in the earliest stages of Alzheimer's disease (5). These soluble oligomers induce acute electrophysiological changes in neurons (6) and are neurotoxic at much lower concentrations than amyloid fibrils (7). Moreover, these soluble oligomers form rapidly in vivo within mildly acidic intracellular compartments and can block long term potentiation, a widely used model for the formation of new memories, at extremely low concentrations (8). Interestingly, non-disease-related proteins can also form highly toxic soluble oligomers, which later form relatively innocuous amyloid fibrils (9). Strikingly, it has been reported that the toxicity of the soluble oligomers of A␤ and other proteins can be blocked by an antibody that binds to them but not to their monomeric or insoluble amyloid forms (10).
To date, at least two soluble neurotoxic A␤ oligomers have been identified and characterized: 1) a large spherical form called "␤-ball" present at pH 2.5 containing about 200 monomers and a high content of ␤ secondary structure (10,11) and 2) smaller ␤-rich spherical or protofibril oligomers formed at neutral pH containing an SDS-resistant core (7,8,10,(12)(13)(14). An additional oligomeric state forming at pH 1 that shares certain similarities with the ␤-balls has been extensively characterized (15).
Structural studies of A␤ oligomers should reveal insights into the biological origins of Alzheimer's disease and may yield new approaches for treatment. The study of A␤ structure at low pH offers the advantage that the ␤-balls do not evolve to form larger aggregates, such as fibrils (11), as occurs at neutral or endosomic pH (4). Moreover, at low pH, the direct measurement of hydrogen exchange rates is possible. One plausible mechanism for A␤ neurotoxicity is the accumulation of the peptide and its oligomers inside the lysosome (pH 4.7), which leads to the rupture of the lysosomal membrane and cell death (16). Therefore, a comparison of the structural tendencies of A␤ at pH 2.5 and neutral pH could give insight into its conformation at the pH that is physiologically significant inside the lysosome. The direct study of the A␤ structures formed at pH 4.7 is very difficult because of the extreme insolubility of the peptide near its isoelectric point (pH 5.5). Finally the ability of an antibody to recognize and block the neurotoxicity of ␤-ball oligomers and the formation of soluble A␤ oligomers at neutral pH is good evidence that they share a structural motif that is key for toxicity (10). For all these reasons, we decided to study the structure of A␤ at low pH.
Our first objective was to characterize the equilibrium between the monomeric and ␤-ball forms of A␤40 using equilibrium sedimentation (ES), CD, and NMR. In addition, the kinetics of ␤-ball formation were studied by fluorescence resonance energy transfer and CD. These experiments provided the data necessary to address the proposal of Zagorski and Barrow (17) that a helical intermediate gives rise to the formation of toxic ␤-sheet rich oligomers of A␤. The data obtained in the present study together with previous results serve as the basis for a plausible model of the ␤-ball structure.
␤-Ball formation could be aided by nascent structure in the A␤ monomer. NMR is an excellent technique for determining the high resolution structure of peptides, but early NMR studies of the A␤ were hindered by the strong tendency of this peptide to aggregate, particularly near pH 5, where it forms a ␤-sheet-rich oligomer that precipitates. This obstacle was first surmounted by the pioneering studies of Zagorski and co-workers, who used cosolvents (18) and detergents (19) to solubilize monomeric A␤. NMR studies of monomeric A␤ in organic solvents (20), in cosolvents (18,21), or associated with detergent micelles (19,(22)(23)(24) revealed extensive helix formation. In contrast, no helical structure was detected in water without cosolvents when more soluble fragments of A␤ peptides and large sample volumes in a 10-mm NMR probe were used to overcome low solubility (25). Recent studies reveal that the full-length A␤ monomer is mostly random coil in aqueous solution near neutral pH, although some segments tend to adopt defined structures (26 -28). Our second goal was to characterize structurally the A␤40 monomer by NMR at low pH in aqueous solution without cosolvents. In addition, the presence of stable secondary structure was tested by direct measurements of hydrogen exchange (HX) that are possible only at low pH. Low peptide concentrations (300 M), detectable with a high sensitivity cryoprobe, were used to avoid precipitation. These measurements revealed further details of the interconversion between monomer and ␤-ball states and established conditions to test for small molecule binding.
A␤40 is positively charged at low pH and has stretches of hydrophobic residues that are likely to be important for the formation of the neurotoxic oligomers. DSS ((CH 3 ) 3 -Si-(CH 2 ) 3 -SO 3 Ϫ Na ϩ ) has recently been reported to bind to basic peptides rich in hydrophobic residues (29). The third objective of this study was to characterize by NMR, equilibrium sedimentation, and fluorescent labeling experiments the ability of DSS to bind to monomeric and ␤-ball forms of A␤40. Moreover the toxicity of DSS and its effect on A␤40 toxicity toward pheochromocytoma (PC-12) cells were determined. On the basis of these data, a plausible, hypothetical structural model for DSS-bound ␤-ball is proposed and is presented here.

EXPERIMENTAL PROCEDURES
Peptide Synthesis-A␤40 was synthesized using the Merrifield solidphase technique with a modified cleavage protocol that minimizes aggregation (11). Guanidine hydrochloride treatment followed by gel filtration at high pH was used to ensure that stock solutions of peptide were completely monomeric (13) and free of oligomers that could act to seed aggregation. A␤40 was stored at 4°C in aqueous solution at pH 10 until use. DSS was obtained from Stohler Isotope Chemicals. All other reagents used were of the highest purity grade commercially available.
Equilibrium Sedimentation-The association state of the A␤40 samples was determined by ES using a Beckman XLI analytical ultracentrifuge. The data were obtained and analyzed as described previously (11).
Circular Dichroism Spectroscopy-Far UV-CD spectra were recorded in a 0.1-cm cuvette at 25.0°C using a JASCO J-810 spectrometer equipped with a Peltier temperature control unit. The scan speed was 10 nm/min, and four accumulations were acquired and averaged to give the final spectrum. Blank spectra were acquired on buffer solutions using identical parameters and subtracted.
The kinetics of ␤-ball formation were followed at 218 or 222 nm and 25.0°C in a 1-mm cuvette with a slit width of 1.3 nm and an instrument averaging time of 8 s. The reaction was started by manually mixing 35 l of A␤40 (pH 10) with 215 l of NaH 2 PO 4 buffer (50 mM). The kinetic dead time was about 50 s. Far UV-CD and UV absorbance spectra were recorded after equilibrium was reached (3-4 h), and the final pH was measured. One-and two-exponential decay functions were fit to the kinetic data using non-linear least-squares algorithms. The uncertainties given are the standard deviation (1) of three trials; the fitting errors from each experiment were larger (about 35%).
Fluorescence Resonance Energy Transfer Experiments-A␤40 peptides were conjugated with a fluorescence donor, Trp, or alternatively to a fluorescence acceptor, ethyldiaminonaphthalene-1-sulfonic acid (EDANS), at the N terminus using a single Gly residue as a flexible spacer as described previously (4). Upon oligomerization of mixed conjugated samples of A␤40, the increased proximity of the EDANS and Trp groups leads to fluorescence resonance energy transfer, producing an increase in the fluorescence emission ratio of EDANS:Trp.
NMR Spectroscopy-To prepare samples for NMR spectroscopy, A␤40 peptide stock solutions were rapidly mixed with an excess of 5 mM NaH 2 PO 4 buffer, pH 2.5. Rapid transfer from alkaline to acidic pH minimizes the time that A␤40 spends near pH 5 where its solubility is minimal. Small pH adjustments were made by adding HCl, DCl, NaOH, or NaOD. All NMR spectra were recorded at 25 33), and natural abundance 13 C heteronuclear single quantum correlation spectra (34) were acquired on a Bruker 600 MHz Avance NMR spectrometer equipped with a triple resonance ( 1 H, 13 C, and 15 N) cryoprobe and Z-gradients. The spectra were assigned using standard methodology (35).
To detect secondary structure, patterns and intensities of NOE backbone signals were examined, and the chemical shift index (36) for 1 H␣, 13 C␣, and 13 C␤ nuclei were determined by applying the procedure and using the reference ␦-values of Wishart et al. (37) except for Asp, Glu, and His in which case the reference ␦-values of Schwarzinger et al. (38) obtained at pH 2 were used. The ⌬␦ profile for N 1 H was based on the reference ␦-values of Wang and Jardetzhy (39).
For HX experiments, A␤40 peptide samples were first concentrated to about 1 mM in pH 2.5, 5 mM NaH 2 PO 4 buffer by ultrafiltration using a YM-3 Amicon filter with a 3-kDa molecular weight cut-off. Then 150 l of this solution was added to 450 l of D 2 PO 4 Ϫ -buffered D 2 O, and HX was monitored by a recording a series of one-dimensional 1 H NMR spectra (dead time ϭ 270 s, 80 transients/spectrum, acquisition time/ spectrum ϭ 120s at 25.0°C). Spectra were recorded until equilibrium was reached (Ϸ70 min). The HX rates were determined by fitting an exponential decay function to the area or height of each peak in the backbone amide region of the one-dimensional 1 H spectra versus time. The observed HX rates were corrected for back-exchange due to 25% 1 H in the solution. The HX experiment was repeated once. Theoretical "k coil " rates for HX from short, unstructured peptides were calculated at the same conditions using the parameters of Bai et al. (40).
Sulfhydryl Rhodamine B Cytotoxicity Assay-PC-12 cells were plated at 1,000 cells/well in a 96-well plate with Dulbecco's modified Eagle's media (Invitrogen) containing 10% fetal bovine serum and incubated overnight. Cells were then differentiated for 4 days in serum-free media containing 20 ng/ml nerve growth factor and N2 supplement. Separately A␤40 (25 M) samples with or without 1 mM DSS were stored for 24 h at pH 5 at room temperature. After differentiation, cells were incubated with the A␤40 samples for 24 h. Toxicity was assayed using the sulfhydryl rhodamine B assay. Briefly, cells were fixed with 10% trichloroacetic acid for 0.5 h, washed with H 2 O, and air-dried. Protein was stained with 0.4% sulfhydryl rhodamine B (Molecular Probes, Inc.) in 1% acetic acid for 0.5 h, washed with 1% acetic acid, and air-dried. The dye was extracted in 10 mM Tris base, and the cell survival, as absorbance at 560 nm, was assayed on a Tecan microtiter plate reader.

A␤40
Exists as Monomers and ␤-Balls at Low pH-The oligomerization state of 300 M A␤40 at pH 2.5, 25.0°C was determined by ES. About half the A␤40 molecules were monomeric, whereas the rest were in large, soluble oligomers whose molecular mass, 764 kDa, corresponds to 176 A␤40 monomers; similar results have been obtained previously (11). A representative CD spectrum of A␤40 in these same conditions shows a broad minimum between 210 and 220 nm characteristic of ␤-structure (Fig. 1A). The magnitude of the minimum, ⌰ 217 ϭ Ϫ2,000 Ϯ 50 degrees cm 2 dmol Ϫ1 , was significant as the signal of a 100% ␤-structure is about Ϫ12,000 degrees cm 2 dmol Ϫ1 (41). Thus, at pH 2.5, 25.0°C, A␤40 forms large oligomers rich in ␤-structure. These oligomers have been characterized previously and named ␤-balls (11).
To determine the kinetics of ␤-ball formation, the pH of a monomeric solution of A␤40 was dropped from 10 to 2.5 (final pH), and the CD signal was followed. Biphasic kinetics with lifetimes () of 1 ϭ 6.9 Ϯ 1.2 min and 2 ϭ 88 Ϯ 4 min were observed, and no "overshoot" indicative of transient helix formation was detected (Fig. 1B). Changes in CD are sensitive to secondary structure formation. As an alternative approach, the kinetics of quaternary structure formation was studied by fluorescence resonance energy transfer. Two solutions with 300 M A␤40, one with untagged A␤40 peptides and the other with A␤40 peptides labeled with Trp or EDANS, were mixed after preincubation at pH 2.5 at 25.0°C. The fluorescence emission decrease with time as unlabeled A␤40 enters and labeled A␤40 exits the initially fully labeled ␤-balls showed biphasic kinetics with 1 ϭ 7 Ϯ 1 min and 2 ϭ 70 Ϯ 40 min (Supplemental Fig.  1). A control CD spectrum revealed that Trp or EDANS labeling did not significantly affect ␤-ball formation (data not shown). Structural Characterization of A␤40 by NMR at Low pH-The region of the two-dimensional 1 H TOCSY spectrum containing the N 1 H to ␣H and side chain correlations of 300 M A␤40 recorded at 25°C at pH 2.5 is shown in Fig. 2A. The linewidth of the ⑀ 1 H of His 14 in D 2 O was about 9 Hz. This signal corresponds to the monomer; those corresponding to the ␤-balls were broadened beyond the detection limit due to their slow correlation time. The monomer linewidth is consistent with slow (Ͼs) interconversion of the monomer and ␤-ball states. The line shapes and intensities of A␤40 NMR resonances remained essentially constant over a period of several months, and the linewidths and chemical shift (␦) values did not change when A␤40 was diluted to 40 M. While the peaks were generally well resolved, the range of N 1 H ␦-values was relatively small (8.74 -8.04 ppm), and Gly ␣H 1 and ␣H 2 signals were essentially degenerate; these are symptoms of random coil backbone structure. No fine splitting of the 1 H peaks was observed, and therefore no vicinal coupling constants were measured. The backbone 1 H assignments were complete with the lone exception of the N 1 H of Asp 1 . The ␦-values of the side chain protons showed little dispersion with variations generally being Յ0.03 ppm for homologous nuclei of the same type of amino acid. The larger ␦-value differences (0.10 -0.27 ppm) observed for Val 12 and Val 18 side chain protons relative to those of other Val residues are likely due to ring current effects from His 13 and Phe 19 , respectively.
Comparison with the assigned 1 H spectra and intrinsic 13 C ␦-values permitted the assignment of all the 13 C␤ and many 13 C side chain and 13 C␣ backbone nuclei in the 13 C heteronuclear single quantum correlation spectra. The assignment of other 13 C␣ nuclei was impeded by their overlap or proximity to the water line. The 1 H side chain assignments are complete except for rapidly exchanging protons. The ␦-values of the ⑀-methyl group of Met 35 , 1 H (2.09 ppm) and 13 C (17.1 ppm), are consistent with its sulfur being unoxidized. The ␦-values of protons in Asp and Glu residues indicate that they are in the neutral, carboxylic acid form, whereas those in His, Lys, and Arg show they are charged. These charged states are consistent with the thorough study of A␤ pK a values reported previously (42). All of the resonances of A␤40 assigned here have been deposited in the BioMagResBank data base (www.bmrb.wisc.edu) under accession number 6257.
In general, strong N 1 H i to C␣ 1 H i Ϫ1 NOEs were observed in the NOESY spectra, whereas N 1 H i to N 1 H i ϩ1 NOEs were generally weaker or absent (Fig. 2B). These relative NOE intensities are characteristic of extended structures such as ␤-strands (35). However, medium-strong NH i to NH i ϩ1 NOEs, which are indicators of helix or turn regions, were observed in residues 12-13, 22-25, and 33-35 (Fig. 2B). NOEs were also observed between the aromatic side chains of Phe 19 and Phe 20 with the alkyl side chains of Val 18 , Ala 21 , and Val 24 (Fig. 2C), indicating that these apolar side chains lie close together and form a hydrophobic cluster, which is at least partially populated. Several different backbone conformations for the turn detected by NOEs between residues 22-25 could position the side chain of Val 24 near those of residues 18 -21. Given the low number of medium and long range NOEs observed no structure calculations were undertaken.
Rules based on the differences between reference and experimental ␦-values, known as the chemical shift index, permit the identification of stable secondary structural elements (36, 43). This approach was applied to the ␦-values of A␤40 obtained here, and no groups of three or four consecutive residues were found to exceed the threshold values; this indicates there is no stable secondary structure in A␤40 in H 2 O at pH 2.5 at 25°C (Fig. 3). In polypeptides, like monomeric A␤40, that lack stable backbone structure, conformational shifts (⌬␦ ϭ ␦ observed Ϫ ␦ reference) can indicate tendencies to adopt partly populated secondary structural elements. According to the ⌬␦ 1 H␣ values, some residues in the N-terminal half of A␤40 appeared to favor the helical conformations, whereas a somewhat stronger trend toward ␤-conformations was observed from Val 24 to the C terminus (Fig. 3A). A general propensity to adopt helix was detected by the ⌬␦ 13 C␣ values, whereas the ⌬␦ 13 C␤ values showed that the central hydrophobic region of A␤40, Leu 17 -Ala 21 , tended to adopt ␤-structure (Fig. 3, B and C). The ⌬␦N 1 H values also showed a general tendency of the polypeptide chain to adopt ␤-structure (Fig. 3D). The 13 C␣ conformational shifts discern helix from random coil better than ␤ from random coil, whereas the opposite is true for the other three nuclei (39). This could account for the different tendencies detected by the ⌬␦ 13 C␣ values. The magnitudes of these shifts were small; a change of 0.5 ppm in ␦ 13 C␣ is one-sixth that observed for protein helices (44), and the ⌬␦ 13 C␤ of residues 17-21 corresponded to a ␤-population Ϸ25% (45). Overall, while these results suggest a tendency for the central and C-terminal nonpolar residues to adopt a small population of ␤-conformations, they clearly indicate the lack of stable secondary structure in these conditions.
Hydrogen Exchange-The HX data could be well fit by a single exponential equation, indicating monophasic kinetics, and essentially identical rates were obtained if peak heights were fit instead of peak areas (data not shown). The final peptide concentration was 0.24 mM. The HX rates are given in Table I. Each peak monitored contained contributions from one to several amide protons. The HX rates from the two experiments agreed within a factor of two. For comparison, the theoretical k coil rates for exchange of unprotected amide protons are listed in Table I resonance in the upfield region of NMR spectra that is frequently used as the ␦ reference. However, when 60 M DSS was present in the A␤40 NMR sample, a broadened signal with a linewidth of 7 Hz, which is similar to those of monomeric A␤40 resonances, was observed at this upfield position, and its ␦-value was provisionally taken as 0.00 ppm (Fig. 4A). This signal was observed in the COSY or TOCSY spectra, but produced no cross-peaks away from the diagonal even when the TOCSY mixing time was increased from 60 to 90 ms (data not shown). These results are compatible with the assignment of the broadened peak to the trimethyl moiety of DSS, which would not give rise to COSY or TOCSY cross-peaks with the methylene protons because the corresponding scalar couplings are negligible. Moreover the 1 H assignments of A␤40 obtained in the present study are essentially complete, and no A␤40 1 H resonated below 0.75 ppm. Meanwhile, in the NOESY spec-trum, the 0.00 ppm peak produced broadened cross-peaks at 0.67, 1.81, and 2.96 ppm (Fig. 4B); these are ␦-values of DSS's methylene groups. To confirm that the broadened signal at 0.00 ppm was due to DSS and did not arise from the peptide, the amount of DSS in the sample was reduced by ultrafiltration, and the NMR spectrum recorded afterward revealed that the broad 0.00 ppm peak was much decreased relative to A␤40 resonances (Fig. 4C). On the basis of these results, we conclude that DSS binds to A␤40.
Addition of a capillary containing DSS to an A␤40 sample allowed the ␦ reference to be established in the absence of binding. This DSS resonated with a narrow linewidth of 2.1 Hz at 0.00 ppm in the same position as the broadened signal observed for A␤40-bound DSS (Fig. 4D). This shows that A␤40 binding does not significantly alter the ␦-values of DSS and suggests that the trimethyl moiety of DSS is not rigidly fixed  19 and Phe 20 , and additional NOEs were detected between DSS and methyl groups that could belong to Leu 17 or Val 18 (Fig. 4, B and E). These data indicate that DSS binds specifically to the central hydrophobic region. In the presence of 200 M A␤40, 60 or 240 M (trimethylsilyl)-propionic acid, which is neutral at pH 2.5, also showed a broadened lines, whereas 1 mM dioxane, which lacks a trimethylsilyl group, shows sharp lines (data not shown).
The interaction between DSS and A␤40 could induce ␦-values changes in both molecules. As a test, the titration of the 200 M A␤40 with 0 -1,000 M DSS was followed by one-dimensional 1 H NMR. As expected, no signal at 0.00 ppm was observed for A␤40 without DSS. No significant DSS-induced changes in the ␦-values or linewidths of the peptide were seen. A broad DSS signal was observed at the lowest DSS concentration tested (20 M), and its linewidth decreased only slightly with increasing DSS (Fig. 4F), which suggests that at 1,000 M DSS more than one DSS molecule is bound per A␤40 monomer. At the end of this titration, a known amount of Trp was added as a concentration standard. By comparing the NMR peak integrals, the concentrations of A␤40 and DSS were found to be 70 and 470 M when their actual concentrations were 120 and 730 M, respectively. Taking the NMR-invisible fraction of A␤40 and DSS as being incorporated into ␤-balls indicates that about 40% of A␤40 molecules were in ␤-balls. The ratio of NMR-invisible DSS:A␤40 was 260 M:50 M, which suggests that about five DSS molecules are bound per A␤40 monomer in ␤-balls.
The Effect of DSS on the A␤40 Monomer to ␤-Ball Equilibrium-Samples of A␤40 peptides (40 M) labeled with EDANS or Trp were incubated overnight with DSS (0 -1,000 M) to test whether DSS binding affects the formation of ␤-balls. A notable fluorescence enhancement was observed in the lowest concentration of DSS tested (100 M) (Fig. 5A). This is good evidence that DSS binds to the A␤40 peptides and promotes their oligomerization. No DSS-induced increase in the amount of ␤-structure in ␤-balls was detected by CD.
To determine whether DSS affects the overall size of the ␤-balls, their mass was determined by ES in the presence or absence of DSS. The number of A␤40 monomers per ␤-ball rose from 260 to 478 in 1.0 mM DSS (Fig. 5B). Thus, the size of the ␤-balls increases remarkably, approximately doubling, in the presence of DSS.
DSS Modulation of A␤40 Cytotoxicity-As a first approach to evaluate the possible in vivo effects of DSS, the sulfhydryl rhodamine B assay was used to test the toxicity of A␤40, DSS, and DSS ϩ A␤40 toward PC-12 cells. DSS was found to be non-toxic at concentrations up to 1 mM and not to affect the toxicity of A␤40 toward PC-12 cells (data not shown).

Monomeric A␤40 Forms ␤-Balls without Apparent Helical
Intermediates-ES and CD clearly detected the presence of large, soluble, neurotoxic ␤-ball oligomers at pH 2.5, 25°C. The molar ratio of A␤40 molecules in the monomeric:␤-ball states was about 1:1 in these conditions. The invariability of the NMR spectrum over several months indicates that the equilibrium between the monomeric and ␤-ball forms is stable once constituted, and no additional oligomers or aggregates are present in significant amounts. Previous work at 4°C at pH 3 also showed that ␤-balls have significant ␤-secondary structure and are stable as they give essentially identical ES results after storage at 0°C for 1 month (11).
The NMR linewidths indicate that the interconversion of monomeric and ␤-ball forms is slow on the time scale of seconds or more. Both fluorescence resonance energy transfer and CD detected biphasic kinetics with similar lifetimes ( 1 ϭ 7 min and 2 ϭ 80 min), which strongly suggests that oligomerization, as detected by fluorescence resonance energy transfer, and acquisition of ␤ structure, as monitored by CD, occur simultaneously. These results also suggest that an oligomeric intermediate with a moderate amount of ␤ structure forms. The population of this intermediate at equilibrium must be low as it was not detected by NMR or ES. The lack of an additional kinetic phase in the HX kinetics indicates that either the amide protons in ␤-balls are unprotected or that the dissociation of ␤-balls into A␤40 monomers has a time constant slower than tens of minutes. The minimal scheme accounting for the kinetic results presented here is below. monomer^intermediate^␤-ball 1 ϭ 7 min 2 ϭ 80 min SCHEME 1 The transient intermediate seems to be "on-pathway" (46) since its oligomerization state and content of ␤-structure are between those of the monomer and ␤-ball states. Although the data do not provide support for the "off-pathway" model of Ref. 17, for the formation of ␤-sheet-rich oligomers via a highly  (17,18,21) or associated with detergent micelles, which mimic some membrane features, (19,(22)(23)(24), may well be due to the ability of these substances to promote helix formation in peptides with a certain innate helix propensity (47,48). Moreover the inability of SDS to disrupt the potently neurotoxic oligomeric conformations of A␤ that disrupt synapse function (7,10) suggests that the helical conformation formed by monomeric A␤ associated with SDS micelles is not akin to pathogenically relevant structures. In contrast, when A␤ peptides interact in vitro with authentic membrane lipids such as gangliosides (49)  glycero-3-phosphocholine/1-palmitoyl-2-oleoyl-sn-glycero-3phosphoglycerol (POPC/POPG) vesicles (50), a rapid induction of fibrils or ␤-structure, respectively, is observed. Gangliosides and cholesterol are present at high concentrations in lipid raft domains of the cell membrane (51), and studies have recently correlated cholesterol levels with in vivo A␤ production and oligomerization (52) and with the risk for developing Alzheimer's disease (53).
Structure Preferences in Monomeric A␤40 Are Similar at Neutral and Low pH-One aim of this work was to characterize the nascent structure in monomeric A␤40 that could play important role(s) in the formation of soluble neurotoxic oligomers. The NMR results showed that the backbone structure of monomeric A␤40 is generally random coil with the central and, perhaps to a lesser extent, C-terminal hydrophobic regions showing some propensity to adopt ␤-structure. Medium-strong backbone NH i to NH i ϩ 1 NOEs revealed turnlike backbone conformations for residues 22-25 and also [33][34][35]. The close similarity of the observed HX rates for A␤40 and the theoretical rates for unprotected HX means that there is no stable secondary structure in monomeric A␤40. The lack of long range NOEs is consistent with an absence of stable tertiary structures. The side chains of Val 18 , Phe 19 , Phe 20 , Ala 21 , and Val 24 form a local hydrophobic cluster as detected by NOEs. The lack of significant ␦-values deviations for these residues, except Val 18 , suggests that the packing of their side chains is loose.
There are notable similarities between the structural propensities detected here for A␤40 in aqueous solution at pH 2.5 and those obtained by others near neutral pH. Working with the fragment A␤-(10 -35), Zhang et al. (26) observed an extensive hydrophobic patch formed by side chains from the central hydrophobic region (Leu 17 -Ala 21 ) together with the side chains of Tyr 10 , Val 12 , and Ile 32 . They described the backbone conformation as a "compact random coil" with residues 22-33 forming a series of turns. The more extended backbone conformation reported here could result from increased charge-charge repulsion in A␤40 at pH 2.5 (Z ϭ ϩ7) versus A␤-(10 -35) at pH 5.7 (Z ϭ ϩ1). There is good evidence that denatured protein chains expand as their net charge increases (54). Working with A␤40 or A␤42 at pH 6.4, Riek et al. (27) reported a defined backbone conformation for residues 16 -24, a helical turn for residues 20 -24, and hydrophobic contacts among the side chains of Leu 17 , Phe 19 , Phe 20 , Ala 21 , and Val 24 with the rest of the peptide being random coil. A comparison of the ␦-value differences of the ␣ 1 H resonances of A␤40 at pH 2.5 and those reported at neutral pH (26,27) revealed a close resemblance that suggests that the backbone conformations are similar at pH 2.5 and neutral pH. Hou et al. (28) also concluded that A␤40 and A␤42 are mostly unfolded at pH 7.2 in water but have two partial populated ␤-strands spanning residues 17-20 and 31-35 and turns encompassing residues 7-11 and 20 -26. The fact that similar structural tendencies were found here at pH 2.5 as near neutral pH corroborates the hypothesis that these nascent structures in monomeric A␤40 represent the first step toward the formation of soluble, neurotoxic oligomers.
DSS Binds to the Central Hydrophobic Cluster of Monomeric A␤40 -The broadened linewidth of the DSS signals and NOEs between DSS and the aromatic rings of Phe 19 and Phe 20 in A␤40 indicate that DSS binds to monomeric A␤40. DSS was reported to bind positively charged, hydrophobic peptides (29); and ␤-cyclodextrin (55) also binds the central hydrophobic region of A␤. The ability of (trimethylsilyl)-propionic acid, which lacks a negative charge at pH 2.5, to bind the hydrophobic segment of A␤ (residues 12-28) (56) suggests that the trimethylsilyl moiety is key for DSS binding to A␤40. Additional NOEs were observed in the present work between the aliphatic methyl groups in A␤40 and the DSS trimethyl moiety. If these methyl groups belonged to Leu 17 or Val 18 the sulfonate moiety of DSS would be positioned to form a favorable electrostatic interaction with the charged amine group of Lys 16  DSS and A␤40 Toxicity Toward PC-12 Cells-There is intense interest in finding molecules that block the formation of neurotoxic, oligomeric species of A␤ peptides (55,57), and here DSS was found to bind to both monomeric and oligomeric A␤40. In an initial series of experiments, DSS was found neither to be toxic nor to affect A␤40 toxicity toward PC-12 cells. Future experiments in more sophisticated model systems, such as nerve tissue or transgenic mice, could further define the potential of DSS to affect A␤ neurotoxicity. Hypothetical Plausible Models for A␤40 ␤-Ball Structure-In developing models for the ␤-ball structure in the presence and absence of DSS, the following structural data were considered. 1) At pH 2.5, all the carboxylate groups, including the Cterminal carboxylate group, are almost completely in the neutral carboxylic acid form. 2) The Arg, Lys, and His residues, as well as the N-terminal amine, are essentially completely charged. These charged groups are concentrated in the Nterminal third of the molecule with the only charged group beyond Lys 16 being Lys 28 (Fig. 6A). 3) The ␤-ball conformation is well populated at pH 2.5-3.0 but is not found above pH 4 where different, insoluble oligomers rich in ␤-secondary structure form and precipitate (11). 4) The range of the number of monomers in the ␤-balls is limited; sizes of 150 -260 monomers per ␤-ball were observed in our Toronto laboratory for samples with slightly different buffer conditions and monomer concentrations. 5) ␤-Balls have a significant ␤-secondary structure as judged by CD. 6) The A␤40 sequence has two regions that are rich in nonpolar residues: Leu 17 -Ala 21 and the last 11-12 residues at the C terminus. 7) These two regions have a slight tendency to adopt extended backbone structures at pH 2.5 and 25°C as revealed by the conformational shifts and backbone NOE intensities in monomeric A␤. 8) A turn, residues 22-25, stabilized by a small hydrophobic cluster involving Val 24 , Ala 21 , Phe 20 , Phe 19 , and Val 18 , detected by NOEs in monomeric A␤, may form between these two hydrophobic regions. It should be borne in mind that the NMR data reported here correspond to the A␤ monomer. 9) ␤-Balls, as resolved by atomic force microscopy, appear spherical and have a diameter that ranges from 8 to 18 nm with a mean of 15 nm (11). 10) The trimethylsilyl and first methylene moieties of DSS bound specifically to the aromatic rings of Phe 19 and Phe 20 and to additional alkyl side chain(s), which most likely belong to the central or Cterminal hydrophobic regions. 11) DSS was found to promote the oligomerization of A␤40 and cause the number of A␤40 monomers in the ␤-balls to double. ␤-Balls with No DSS-Based on these data, we propose that ␤-balls have a spherical micelle structure with the C-terminal 12 residues buried inside a hydrophobic micelle core and the rest of the peptide, including the charged Lys 28 , exposed to solvent (Fig. 6B). The peptide groups within the ␤-ball core are hydrogen-bonded forming ␤-sheet structures. In an extended conformation, these last 12 residues of A␤40 would be about 4.1 nm long; therefore the radius of the proposed ␤-ball model corresponds well with its minimal diameter detected by atomic force microscopy (8 nm). With the last 12 residues of A␤40 modeled as a cylinder about 4.1 nm long with a radius of 0.25 nm, it can be estimated that about 350 A␤40 C-terminal tails could fit into the micelle core.
It should be kept in mind that the ␤-ball model proposed here is hypothetical. Other geometric forms, like elliptical or cylindrical micelles or bilayers, also permit the solvent exposure of a polar head group and the burial of a nonpolar tail. Adapting the approach of Tanford (58) and approximating the last 12 residues of A␤40 as a cylinder, the surface area available to Lys 28 , the first exposed residue, would be ϳ60 Å 2 in a spherical micelle, 50 Å 2 in a cylindrical micelle, and just 20 Å 2 in a bilayer. Thus, the positive charges in the N termini will be the farthest apart and the most solvent-exposed in a spherical micelle. Both cylindrical micelles and bilayers can grow indefinitely on their ends or edges, respectively, whereas the number of monomers in spherical or elliptical micelles is limited geometrically. Therefore, the observation that the size distribution of ␤-balls is limited is more consistent with spherical or elliptical micelle geometries. ␤-Balls have not been detected above pH 4. Our model predicts that ␤-balls will be unstable above this pH since titration of the C-terminal carboxylate group (pK a Ϸ 3.8) favors its solvent exposure and will break any hydrogen bonds that may form between the neutral carboxylic acid groups and disrupt the hydrophobic ␤-ball core. A spherocylindrical A␤ micelle forming at pH 1 has been thoroughly studied by Yong et al. (15). Like the ␤-ball, the pH 1 oligomer could contain buried Cterminal carboxylic acid groups and has a definite size range. Relative to the ␤-ball, the pH 1 spherocylindrical micelle contains fewer monomers (30 -50), and its dimensions suggest that the nonpolar residues in the both the central and C-terminal hydrophobic segments are buried within the micelle core. The higher ionic strength present at pH 1 could more effectively screen the charge-charge repulsions allowing Lys 28 from different monomers to be positioned closer together and thus account for its different geometry compared with ␤-balls. Other spherical A␤ peptide oligomers are present at endosomic (4) or neutral pH (7,(12)(13)(14) prior to the formation of amyloid fibrils or protofibrils (Table II). The smaller diameter of many of these neutral pH oligomers is consistent with the finding that their C-terminal carboxylate groups are not buried but are charged and exposed. In contrast to the low pH A␤ oligomers, at neutral pH both positively and negatively charged groups are present and could be arranged to permit favorable electrostatic interactions in the oligomers (59). The positioning of multiple positive charges on the exterior of the ␤-balls probably keeps them well separated in solution and impedes their evolution to fibrils. In contrast, favorable electrostatic interactions in neutral pH spherical oligomers likely permit them to form progressively higher order aggregates. The existence of a common structural motif in both the ␤-balls, which lack favorable electrostatic interactions, and the neutral pH oligomers, wherein favorable electrostatic interactions are likely, that is specifically recognized by a neurotoxicity-neutralizing antibody (10) suggests that electrostatic interactions are not key for the stability or formation of this structural motif.
␤-Balls with DSS-In the presence of DSS, the number of A␤40 monomers in the ␤-balls approximately doubled to about 476 monomers. This number of monomers is too large to fit into the hydrophobic core of the micelle but could be accommodated as a second shell (Fig. 6C). In this plausible, hypothetical model, the C-terminal and central hydrophobic regions of each additional molecule are proposed to interact with the central hydrophobic regions of two adjacent A␤40 molecules whose C termini are buried in the micelle core (Fig. 6C). DSS would promote the oligomerization of A␤40 monomers by binding to hydrophobic regions and Lys residues bearing positive charges. The strongest DSS binding site is proposed to be the central hydrophobic regions and positively charged Lys 16 residues of two A␤40 monomers, one each from the inner and outer shells (Fig. 6C). Other plausible DSS binding sites are indicated. New NMR techniques like transverse relaxation-optimized spectroscopy (60) and cross relaxation-induced polarization transfer (61) yield good quality spectra of protein complexes up to 900 kDa at high magnetic field strength; these methods will be applied to study the high resolution structures of free and DSS-bound ␤-balls.