Kinetic Studies of Amyloid β-Protein Fibril Assembly

Amyloid β-protein (Aβ) fibril assembly is a defining characteristic of Alzheimer's disease. Fibril formation is a complex nucleation-dependent polymerization process characterized in vitro by an initial lag phase. To a significant degree, this phase is a consequence of the energy barrier that must be overcome in order for Aβ monomers to fold and oligomerize into fibril nuclei. Here we show that low concentrations of 2,2,2-trifluoroethanol (TFE) convert predominately unstructured Aβ monomers into partially ordered, quasistable conformers. Surprisingly, this results in a temporal decrease in the lag phase for fibril formation and a significant increase in the rate of fibril elongation. The TFE effect is concentration dependent and is maximal at ∼20% (v/v). In the presence of low concentrations of TFE, fibril formation is observed in Aβ samples at nanomolar concentration, well below the critical concentration for Aβ fibril formation in the absence of TFE. As the amount of TFE is increased above 20%, helix content progressively rises to ∼80%, a change paralleled first by a decrease in elongation rate and then by a complete cessation of fibril growth. These findings are consistent with the hypothesis that a partially folded helix-containing conformer is an intermediate in Aβ fibril assembly. The requirement that Aβ partially folds in order to assemble into fibrils contrasts with the mechanism of amyloidogenesis of natively folded proteins such as transthyretin and lysozyme, in which partial unfolding is a prerequisite. Our results suggest that in vivo, factors that affect helix formation and stability will have significant effects on the kinetics of Aβ fibril formation.

The amyloidoses are a group of disorders characterized by aberrant protein folding and assembly, leading to the deposition of insoluble protein fibrils, cell and organ dysfunction, and in many cases, death (1)(2)(3). At least 18 different proteins and peptides form amyloid deposits in vivo (4). Though these proteins are non-homologous and have diverse native tertiary and quaternary structures, all polymerize into extended fibrils with diameters of ϳ7-10 nm that are stained by amyloidophilic dyes such as Congo red and Thioflavin S. The dye-binding property of these assemblies is a result of the high ␤-sheet content within the fibrils (5). X-ray diffraction analysis has revealed a cross-␤ structure, i.e. the ␤-strands are oriented perpendicular to the fibril axis and stack into axially aligned extended ␤-sheets (6,7). The underlying mechanism of fibril formation for several of the amyloidogenic proteins, including lysozyme and transthyretin, involves a partial destabilization of the protein's native conformation through either mutations or environmental changes (8,9). The resulting conformer is prone to aggregation, and its native-like secondary structure is converted to an extended ␤-sheet structure during this process. In addition, certain non-amyloidogenic proteins can form fibrils when partially denatured by incubation at acidic pH or in the presence of organic solvents (10 -12).
In contrast to fibrillogenesis of natively folded proteins, amyloid ␤-protein (A␤) 1 fibril formation in vitro involves the conversion of an irregularly structured conformer into a highly stable, ␤-sheet-rich assembly (for review, see Ref. 13). In vivo, the structure of the nascent A␤ monomer is unknown. 〈␤ is produced through successive endoproteolytic cleavage events adjacent to and within the transmembrane domain of the amyloid ␤-protein precursor (A␤PP) (14). The structure of the extramembranous portion of the A␤ region of A␤PP, comprising the N-terminal 28 amino acids of A␤, has not been determined. The remaining 12-14 amino acids composing the C-terminal portion of A␤ may exist as a helix in A␤PP; however, peptide bond hydrolysis involves helix disruption (15). Both random coil (RC)3␤-sheet and ␣-helix3␤-sheet transitions thus could occur during A␤ folding and assembly.
Recent in vitro studies of A␤ fibrillogenesis support the hypothesis that helix 3 strand transitions play a prominent role in the fibril assembly process. Walsh et al. (16,17) used circular dichroism (CD) spectroscopy to study conformational changes occurring during protofibril and fibril formation by A␤ . These studies suggested that fibrillogenesis involved the transient formation of a helix-containing intermediate (16). In an extensive series of experiments examining the fibrillogenesis of 18 clinically relevant alloforms of A␤, Kirkitadze et al. (18) confirmed and extended these findings, observing in each case the formation of an oligomeric, helix-containing fibril intermediate. Interestingly, Kallberg et al. (19) have found that several amyloid-forming proteins contain ␣-helices in polypeptide segments, which are predicted, based on theoretical considerations, to form ␤-strands. Based on this finding, it was postu-lated that ␣-helix/␤-strand discordant sequences predict which proteins have a propensity to form amyloid. Both the prion protein and A␤ display this discordance, in agreement with the theory. In addition, direct experimental study of a number of other discordant helix-containing proteins revealed for the first time that they could indeed form typical ␤-sheet-containing fibrils (19). A reasonable postulation, based on these and other data, is that stabilization of ␣-helical conformation could block fibril formation (19). In fact, in the presence of high concentrations (Ͼ40%) of the co-solvent trifluoroethanol (TFE), both A␤(1-40) and A␤  are structured, exhibit predominantly ␣-helical conformations, and do not fibrillize (20 -23). Here, we took advantage of the ability of TFE to facilitate helix formation to examine in greater detail the effect of helix stabilization on A␤ assembly. This approach also has been used successfully to study the effect of helix stabilization on the folding of globular proteins (24). Surprisingly, our results reveal that helix stabilization may facilitate as well as inhibit fibril formation, depending on its strength.

EXPERIMENTAL PROCEDURES
Reagents and Chemicals-Chemicals were obtained from Sigma (Saint Louis, MO) and were of the highest purity available. Water was de-ionized and filtered using a Milli-Q system (Millipore Corp., Bedford, MA).
Peptide Synthesis-A␤  and A␤(1-42) were synthesized using 9-fluorenylmethoxycarbonyl (FMOC) chemistry and purified by reverse phase high performance liquid chromatography (RP-HPLC), essentially as described (17). The identity and purity of the peptides were confirmed by amino acid analysis, mass spectroscopy, and RP-HPLC. Purity levels generally exceeded 97%.
Isolation of Low Molecular Weight (LMW) A␤-The term "LMW A␤" was defined originally as monomeric or dimeric A␤ (16). To isolate LMW A␤, A␤  or A␤  were dissolved at a concentration of 2 mg/ml in 10 mM NH 4 OH, pH 10.8, containing 6 M guanidine hydrochloride and sonicated in a Branson 1200 ultrasonic water bath for 1 min. The solution was centrifuged at 16,000 ϫ g for 10 min. Then, ϳ200 l of the supernate were injected onto a Superdex 75 HR 10/30 column (Amersham Biosciences, Piscataway, NJ) attached to a Waters 650 Advanced Protein Purification System. The column was eluted with 2.5 mM NH 4 OH, pH 10.8, containing 75 mM NaCl, at a flow rate of 0.5 ml/min. Peptides were detected by UV absorbance at 254 nm, and 400-l fractions were collected during elution of the LMW A␤ peak. Published work has shown that fractions prepared in this manner are free of pre-existing aggregates (16,17,(25)(26)(27). The LMW A␤ fraction was prepared for study by dilution into phosphate buffered saline (PBS), pH 7.4, to produce a final solution containing 10 mM buffer and 75 mM NaCl, with the appropriate percentage of TFE. For chromatography at neutral pH, the peptides were prepared as above, except at a concentration of 0.5 mg/ml, and the column was eluted with 10 mM sodium phosphate, pH 7.4, containing 75 mM NaCl and 0% or 20% TFE. Unless otherwise indicated, final peptide concentrations used in the experiments were 30 M (A␤(1-40)) and 10 M (A␤(1-42)).
Circular Dichroism Spectroscopy-LMW A␤ samples were prepared as described above and were either examined immediately after preparation (Figs. 1 and 5) or were incubated prior to CD measurement (Fig.  2). The samples were examined using 1-mm cuvettes (Hellma, Forest Hills, NY) in an Aviv Model 62A DS spectropolarimeter (Aviv Associates, Lakewood, NJ). Spectra were recorded at 25°C from ϳ195-240 nm at 1 nm resolution with a scan rate of 0.25 nm/s. Four scans were acquired and averaged for each sample. Raw data were manipulated by smoothing and subtraction of buffer spectra according to the manufacturer's instructions. Deconvolution of the resulting spectra was achieved using the LINCOMB program (28) and the basis set of Brahms and Brahms (29). The relative amounts of random coil, ␣-helix, ␤-sheet, and ␤-turn were determined from the normalized contribution of each secondary structure element function to the observed spectrum follow-ing curve fitting. Guanidine (Gu⅐HCl) unfolding curves were obtained by monitoring ellipticity at 222 nm. The Gibbs free energy of unfolding (⌬G 0 ) for the unfolding was calculated from the Gu⅐HCl denaturation data according to the method of Santoro and Bolen (30,31). This method applies a non-linear least-squares fit to a linear extrapolation method that evaluates the unfolding free energy at zero denaturant concentration. The linear extrapolation method was originally based on empirical observation of the linear dependence of observed unfolding free energy changes as a function of denaturant (31).

Effects of TFE on A␤ Secondary
Structure-Prior to determining the effect of helix formation on the rate of A␤ fibrillogenesis, we first quantified the effect of TFE on the conformation of A␤ prior to initiation of the fibril formation process. To do so, size exclusion chromatography was used to prepare solutions of fibril-free low molecular weight (LMW) A␤. The term "LMW A␤" was originally defined as monomeric or dimeric A␤ (16). Immediately after collection, these LMW A␤ preparations were diluted into PBS containing various amounts of TFE and then monitored by far-UV circular dichroism. The resulting CD spectra ( Fig. 1, A and B) were then deconvoluted to obtain estimates of the relative amounts of each secondary structure element present (Table I). For both peptides, increasing concentrations of TFE resulted in a conformational transition from a largely disordered structure to a highly helical structure. The initial random coil content of each peptide was ϳ70%, whereas at high TFE concentrations, the ␣-helix content reached ϳ80%. In 20% TFE, intermediate levels of helix content were observed (54% for A␤  and 45% for A␤(1-42)). A plot of the TFE concentration dependence of helix content (Fig. 1C) produces a sigmoidal function with a steep transitional zone between ϳ5-30%. The midpoint of this transition occurs at ϳ15-20% TFE. Maximum helix content (ϳ80%) occurred at a TFE concentration of ϳ40% for A␤(1-40) and ϳ60% for A␤ .
TFE Concentration Dependence of ␤-Sheet Formation-A␤ fibril formation is associated with the formation of ␤-sheets. To determine the effect of TFE on the kinetics of ␤-sheet formation, temporal changes in the secondary structure of A␤ were monitored in the presence of different amounts of TFE. In addition, electron microscopy was used to show that the ␤-sheet-containing assemblies monitored by CD were fibrillar (see following two subsections). When either A␤(1-40) or A␤(1-42) was incubated in 10 -20% TFE, the rate of ␤-sheet formation increased significantly relative to the rate observed in the absence of TFE (Fig. 2, A and B). Interestingly, this effect was not linearly correlated with TFE concentration, but displayed a maximum at ϳ15-20% TFE. An inverse relationship was observed between ␤-sheet and ␣-helix content. For example, at 20% TFE concentration, a 10365% increase in ␤-sheet content was associated with a 50310% decrease in ␣-helix content (data not shown). This type of ␣-helix3␤-sheet transition has been observed previously (18). As the TFE concentration was increased from 25 to 40%, the rate of ␤-sheet formation decreased. At higher concentrations, Ն40% for A␤  and Ն60% for A␤ , both peptides exhibited an elevated and constant helical content of ϳ80%, did not bind Congo red, and produced transparent solutions with no evidence of aggregation, even after 1 month of incubation (data not shown). These results are consistent with NMR studies showing that A␤ is monomeric in 40% TFE (23). The rates of ␤-sheet and fibril formation by A␤(1Ϫ42) were significantly faster than those of A␤(1Ϫ40). However, the qualitative effect of TFE on these rates was similar for both peptides in that the maximum rates were observed with ϳ20% TFE (Fig. 2B). It should be noted that the conclusions regarding TFE-induced rate differences in ␤-sheet formation are the same if rates are calculated as time-to-maximal ␤-sheet content or as time-to-half-maximal ␤-sheet content.
Morphology of A␤ Assemblies Formed in the Presence of TFE-An assumption inherent in the kinetics studies discussed previously was that the conformational changes monitored by CD were associated with fibril formation. To prove that the assumption was true, electron microscopy was used to determine the morphologies of the end-stage assemblies formed by A␤(1-40) and A␤(1-42) following incubation in 0 or 20% TFE (Fig. 3). In the absence of TFE, A␤(1-40) formed long fibrils with diameters of 6 -10 nm (Fig. 3A). These fibrils displayed a regular helical pitch of 60 -80 nm, had smooth margins, and appeared to be composed of two filaments, each 3-3.5 nm in diameter. In 20% TFE, A␤(1-40) also formed long twisted fibrils (Fig. 3B). These fibrils often contained three filaments. A␤  formed both short and long, straight, 6 -10-nm diameter fibrils in the absence of TFE (Fig. 3C). Many of these fibrils had little discernible substructure, whereas others appeared to be twisted and bifilar. In 20% TFE, long, straight fibrils with pronounced helical winding were observed (Fig. 3D). These fibrils appeared to be bifilar in nature. Another prominent feature of the TFE-treated samples was abundant, long, flexible filaments that displayed little substructure. These flexible filaments were often associated with the straight fibrils and in some areas appeared to be wound into them. The morphologic differences observed in the presence and absence of TFE may result from its effect on interchain packing. TFEmediated affects on close packing would be consistent with the ability of TFE to affect hydrophobic interactions (32). Of interest was the fact that fibril formation occurred at much lower concentrations of peptide in the presence of TFE. For example, in the presence of 20% TFE, fibrils were detected in solutions of 500 nM A␤(1Ϫ40) and 200 nM A␤(1Ϫ42) after 18 and 10 days, respectively; whereas at the same peptide concentrations, no fibrils were observed in the absence of TFE even after 1 month of incubation (data not shown).
Aggregation State of A␤ in 20% TFE-To determine whether the TFE-mediated increase in the helical content in A␤ resulted from a direct effect on the A␤ monomer or from an indirect effect associated with A␤ oligomerization, size exclusion chromatography (SEC) was used to characterize the assembly state of A␤. Both in the presence and absence of 20% TFE, A␤(1Ϫ40) chromatographed as a single peak with an identical retention time, a characteristic of LMW A␤ (Fig. 4A). The elution profile of A␤(1-42) also was unaffected by TFE (Fig. 4B). CD spectroscopy was also used to check for aggregation of A␤ in 20% TFE. If the helical structure formed as a result of TFE treatment arose from the association or aggregation of A␤, then it would be expected that the CD signals would be dependent upon peptide concentration. To examine this issue, LMW A␤ was isolated by SEC, diluted into PBS containing 20% of TFE, and then immediately monitored by CD. The data revealed that no change in the CD spectra for A␤(1Ϫ40) occurred over the concentration range 0.8Ϫ70 M, nor were any changes observed for A␤(1Ϫ42) in the range 0.5-10 M (Fig. 5). These results suggest that initially, in these concentration ranges, the LMW A␤ peptides do not aggregate and that the observed increases in helix content occur within the context of the A␤ monomer. A significant decrease in the percentage of helix is observed for A␤(1Ϫ42) at concentrations exceeding 15 M. Helix levels reach a minimum value of ϳ20% at concentrations Ն25 M. This decrease in helix content is associated with a significant increase in ␤-structure content and the rapid formation of fibrils (data not shown). These results suggest that A␤(1Ϫ42) in 20% TFE aggregates rapidly at concentrations Ն15 M. For this reason, A␤(1Ϫ42) was studied at concentrations Յ10 M.
Gu⅐HCl Unfolding of A␤ in the Presence of TFE-The TFE effects observed in our experiments were likely caused by the known ability of this co-solvent to stabilize helices. However, to determine quantitatively the significance of any helix stabilization for A␤ assembly, we compared the free energy of unfolding (⌬G 0 ) of partially helical A␤ (20% TFE) to that of predominately helical A␤ (40% TFE for A␤(1Ϫ40) or 60% TFE for A␤(1Ϫ42)). The molar ellipticity of A␤ at 222 nm was measured as a function of Gu⅐HCl concentration (Fig. 6). The unfolding transitions of A␤  or A␤(1-42) (Fig. 6, open symbols) were sigmoidal and exhibited shallower transitions at 20% TFE than at 40% (A␤(1Ϫ40)) or 60% (A␤(1Ϫ42)). The sigmoidal transition between folded and unfolded states prompted us to analyze these data using two-state transition theory. Although the unfolding of a peptide such as A␤ may not actually be two-state in nature, limits on the stability of the peptide can be derived from this analysis. Reversibility of unfolding was studied by first unfolding A␤ with 6 M Gu⅐HCl and then diluting the predominately unfolded peptide to final Gu⅐HCl concentrations within the transition region (Fig. 6, (ϩ) plus signs). The data were then analyzed by the linear extrapolation method of Santoro and Bolen (30,31) (Table II). The results showed that the conformers formed at high concentrations of TFE by A␤  and A␤(1-42) were significantly more stable than those formed at low TFE concentrations. The "trapping" of these conformers in an energy well explains why high TFE concentrations prevent fibril formation. DISCUSSION Increasing evidence suggests that a seminal event in the amyloidogenesis of most proteins and peptides is the partial unfolding of the native conformer (8,9,33). This unfolding  (29). The percentage of each secondary structure element is listed. "␤-structure" includes both ␤-sheet and ␤-turn. Each experiment was repeated three times, yielding similar data in each case. Results are presented from one such experiment.
A very stable native fold precludes sufficient population of the amyloidogenic state and results in the absence of significant fibril nucleation and elongation.
In contrast to studies of natively folded proteins, in vitro studies of A␤ fibrillogenesis have suggested that fibril assembly involves a transition of the natively unfolded peptide monomer to a ␤-sheetϪrich polymer (13). Recently, careful examination of the conformational transitions associated with the formation of A␤ protofibrils (16) and fibrils (18)  The solid lines are the curves derived following fitting of the data by the method of Santoro and Bolen (30,31). Open symbols signify Gu⅐HClinduced unfolding. Plus (ϩ) symbols indicate refolding. The data shown are representative of those obtained in each of three independent experiments.

TABLE II
Free energies of unfolding (⌬G 0 ) for A␤ Gu⅐HCl-mediated unfolding/refolding of A␤(1-40) and A␤  were monitored by CD. The data were then analyzed by the linear extrapolation method of Santoro and Bolen (30,31) (18). Conversion of the intermediate into fibrils results in the loss of helical elements. Here, to better understand the role of helix formation in the earliest phases of A␤ folding and assembly, the behavior of LMW A␤ was studied in the presence of the helix-facilitating co-solvent TFE. At low concentrations (Ͻ20% v/v), rather than inhibiting fibrillogenesis, TFE significantly accelerated fibril assembly in a concentration-dependent manner. These data support the hypothesis that, in some cases, amyloid fibril formation by natively unfolded proteins requires the "partial folding" of the disordered monomer to form a discrete intermediate, which then undergoes additional conformational transitions in the process of fibril assembly. Recent studies of another natively unstructured peptide, ␣-synuclein, are also consistent with this mechanism (35). This partial folding hypothesis is the mirror image of that invoked to explain the fibrillogenesis of natively folded proteins such as transthyretin (36) or lysozyme (8). TFE is a co-solvent known to destabilize hydrophobic interactions within polypeptide chains and to stabilize local hydrogen bonds between residues close in the amino acid sequence, particularly those forming ␣-helices and ␤-hairpins (37). Because hydrophobic interactions play an important role in A␤ fibril formation, we reasoned that TFE should weaken these interactions and thus decrease the rate of fibril formation. In addition, we expected that stabilization of ␣-helices would slow down or prevent fibril assembly. Surprisingly, although low concentrations of TFE did in fact stabilize ␣-helices in A␤, the rate of fibril formation increased substantially relative to that observed in the absence of TFE. During fibril assembly, the expected increase in ␤-sheet level was mirrored by a proportionate decrease in ␣-helix level. Interestingly, the accelerating effect of TFE was not correlated linearly with its concentration, but exhibited a maximum at ϳ20%. At higher concentrations, the rate of fibril formation decreased progressively, eventually reaching zero. In the presence of 20% TFE, a significant amount of helix existed in A␤ (54% in A␤  and 45% in A␤ ), yet the rate of fibril formation was significantly faster than that observed in conformers displaying maximum random coil content (70% RC) or maximum helix content (80% helix). These results emphasize the importance of the partially helical conformer of A␤ in controlling the kinetics of fibril formation.
An important question is why disordered A␤ monomers form fibrils at a significantly slower rate than when they exist as partially folded, helix-containing conformers. One explanation is that the disordered A␤ conformers are not able to associate to form the stable, oligomeric assemblies required to nucleate fibril formation. Rather, they form a diverse population of fibril-incompetent, unstable aggregates (38). Consistent with this idea, recent studies of the oligomerization state of LMW A␤ prior to fibril nucleation and growth have shown that the peptide exists in a rapid equilibrium involving monomers, dimers, trimers, and tetramers (26). It is likely that partial folding of the A␤ monomer is necessary to facilitate the correct intermolecular packing within A␤ oligomers. These packing interactions provide sufficient energy to overcome the entropic costs of the partial ordering of the A␤ monomer which takes place during this oligomerization. The helical components of these ordered assemblies would then convert into ␤-strands and ␤-turns coincident with nucleus or "seed" formation and then fibril formation. This mechanism is supported by recent studies of the fibrillogenesis of a model 38-residue helix-turn-helix peptide, ␣t␣, which revealed that helix-rich oligomeric intermediates form prior to the appearance of ␤-sheet-rich fibrils (39).
A consideration of basic aspects of protein folding leads to mechanistic interpretations similar to those above. For example, it has been suggested that efficient protein folding can be induced by restricting the conformational space associated with amyloid-incompetent conformers (33). In doing so, the stochastic process by which the natively unfolded conformer searches conformational space as it folds and assembles is less likely to produce off-pathway conformers. Here, helix formation facilitates the organization of nascent disordered A␤ into a partially folded conformer, one that has a high probability of accessing later amyloid-competent conformational states. This results in the efficient movement of A␤ down the pathway of fibril formation and explains the fact that fibril formation is observed even in A␤ samples of nanomolar concentration, well below the critical concentration for A␤ fibril formation in the absence of TFE. It is important to emphasize that A␤ fibril formation may involve more than one pathway. Energy landscape models of protein folding can be exceptionally complex, but do include pathways for fast folding (40,41). The helixcontaining intermediate may be part of a fast folding pathway through which random coil A␤ may form fibrils more rapidly than through other pathways. Other, as yet unidentified, intermediates also may exist. Protein folding theory also predicts that if the partially folded intermediate is stabilized sufficiently, the activation barrier for its structural reorganization into the fibril will be large enough to stop fibril formation (24,42,43). Consistent with this prediction, the further stabilization of helical elements within A␤ obtained using high concentrations of TFE prevented fibril formation. Quantitative guanidine denaturation/renaturation experiments showed that the effects of TFE were associated with significant changes in helix stability. This finding is in agreement with results of fundamental studies of the mechanism of TFE-mediated helix induction in peptides (44) and proteins (24). In the latter work, low concentrations of TFE or hexafluoroisopropanol significantly increased the folding rate of the protein acylphosphatase and converted an apparent two-state folding process into a process characterized by the accumulation of partially structured species. However, at higher TFE concentration, the folding process was inhibited by the formation of additional non-native hydrogen bonds (24). Here, in the A␤ case, low concentrations of TFE accelerated fibrillogenesis through stabilization of local helical structure. This suggests that at early stages in the peptide assembly process, local hydrogen bonding is very important. It is also important to note, as discussed previously, that high concentrations of TFE significantly weaken hydrophobic interactions (37). Thus, in addition to its demonstrated effect on ␣-helix content and stability, the mechanism of inhibition of fibril formation at high TFE concentration could include destabilization of hydrophobic interactions.
Our results lead to certain predictions about the behavior of A␤ in vivo, where the structure of the nascent peptide is unknown. If the peptide is largely helical, which it may be because of its partial composition of the A␤PP transmembrane domain, then significant relaxation of portions of the helix would be necessary for A␤ to fold into an amyloid-competent conformation. Alternatively, A␤ may be largely unstructured, as it is in vitro, because of the fact that it has no prosthetic groups or disulfide bonds to stabilize a folded conformation. Formation of the partially folded intermediate thus would be necessary for fibrillogenesis to occur. In either case, conditions facilitating the population of the partially folded state would be expected to accelerate fibrillogenesis and to allow the process to occur at lower nominal A␤ concentration.
In vivo, A␤(1-40) and A␤(1-42) appear to be seminal players in the early stages of amyloid formation (13). Our examination of the effects of TFE on the folding and assembly of these peptides revealed that each is affected in a qualitatively similar way. TFE facilitates helix formation and accelerates the fibrillogenesis of both peptides in a concentration-dependent manner in which maximal effects are observed at a TFE concentration of ϳ20%. Where and how within the A␤ peptide might the initial coil 3 helix transition occur? Evidence suggests that the answer to the first question is the region of A␤ encompassing Lys 16 -Asp 23 . This octapeptide segment comprises amino acids with high helical propensities (45). Prior conformational studies of A␤ have shown that this region forms helices under the appropriate conditions (23). In addition, subtle changes in amino acid sequence (e.g. Val 18 3 Ala) significantly facilitate nascent helix formation (46), and recent studies have shown directly that helix destabilization in the Leu 17 -Val 24 region leads to fibril formation (47). Finally, the central hydrophobic cluster of A␤, Leu 17 -Ala 21 , which plays a critical role in organizing A␤ folding and fibril formation (48 -50), is contained within the Lys 16 -Asp 23 sequence.
We postulate that intramolecular helix formation occurs as a natural consequence of the sampling of conformational space by the A␤ monomer. Once formed, intermolecular interactions among partially folded, helix-containing A␤ monomers could lead to peptide oligomerization followed by conformational reorganization to form the extended ␤-sheets that compose the mature amyloid fibril. This process is consistent with data obtained in studies of the fibrillogenesis of A␤ alloforms associated with familial forms of cerebral amyloid angiopathy (18) and has been shown to occur in the fibrillogenesis of the model peptide ␣t␣ (39).
In summary, our findings are consistent with the hypothesis that the transition from nascent, natively unfolded A␤ monomer to amyloid fibril involves formation of a partially folded, helix-containing intermediate. The process of structural organization of A␤ into a partially folded intermediate contrasts with the mechanism of amyloidogenesis of natively folded proteins in which partial unfolding is a prerequisite. These results suggest that factors that affect helix formation and stability in vivo, e.g. membranes and lipoproteins, will have a significant effect on the kinetics of A␤ fibril formation. For this reason, therapeutic strategies focusing on helix stabilization must be pursued with care.