Early Synergy between A (cid:1) 42 and Oxidatively Damaged Membranes in Promoting Amyloid Fibril Formation by A (cid:1) 40*

Oxidative lipid membrane damage is known to promote the misfolding of A (cid:1) 42 into pathological (cid:1) structure. In fully developed senile plaques of Alzheimer’s disease, however, it is the shorter and more soluble amyloid (cid:1) protein, A (cid:1) 40, that predominates. To investigate the role of oxidative membrane damage in the misfolding of A (cid:1) 40, we have examined its interaction with supported lipid monolayer membranes using internal reflection infrared spectroscopy. Oxidatively damaged lipids modestly increased A (cid:1) 40 accumulation, with adsorption kinetics and a conformation that are distinct from that of A (cid:1) 42. In stark contrast, pretreatment of oxidatively damaged monolayer membranes with A (cid:1) 42 vigorously promoted A (cid:1) 40 accumulation and misfolding. Pretreatment of saturated or undamaged membranes with A (cid:1) 42 had no such effect. Parallel studies of lipid bilayer vesicles using a dye binding assay to detect fibril formation and electron microscopy to examine morphology demonstrated that A (cid:1) 42 pretreatment of oxidatively damaged membranes promoted the formation of mature A (cid:1) 40 amyloid fibrils. We conclude that oxidative membrane damage and A (cid:1) 42 act synergistically at an early stage to promote

Oxidative lipid membrane damage is known to promote the misfolding of A␤42 into pathological ␤ structure. In fully developed senile plaques of Alzheimer's disease, however, it is the shorter and more soluble amyloid ␤ protein, A␤40, that predominates. To investigate the role of oxidative membrane damage in the misfolding of A␤40, we have examined its interaction with supported lipid monolayer membranes using internal reflection infrared spectroscopy. Oxidatively damaged lipids modestly increased A␤40 accumulation, with adsorption kinetics and a conformation that are distinct from that of A␤42. In stark contrast, pretreatment of oxidatively damaged monolayer membranes with A␤42 vigorously promoted A␤40 accumulation and misfolding. Pretreatment of saturated or undamaged membranes with A␤42 had no such effect. Parallel studies of lipid bilayer vesicles using a dye binding assay to detect fibril formation and electron microscopy to examine morphology demonstrated that A␤42 pretreatment of oxidatively damaged membranes promoted the formation of mature A␤40 amyloid fibrils. We conclude that oxidative membrane damage and A␤42 act synergistically at an early stage to promote fibril formation by A␤40. This synergy could be detected within minutes using internal reflection spectroscopy, whereas a dyebinding assay required several days and much higher protein concentrations to demonstrate this synergy.
Extracellular amyloid deposits known as senile or neuritic amyloid plaques are one of the defining histopathological features of Alzheimer's disease (AD). 1 The core of a neuritic plaque is a compact heterogeneous meshwork in which fibrillar forms of amyloid ␤ (A␤) proteins predominate. The best characterized A␤ proteins are 40 and 42 residues in length, varying at their COOH terminus. The 40-residue A␤ protein (A␤40) and various amino-and carboxyl-terminal derivatives appear to be the pre-dominant A␤ protein in the amyloid plaques of AD (1), in cerebrovascular amyloid deposits (2), and in at least some forms of hereditary amyloidosis (3). The 42-residue A␤ protein (A␤42) appears to associate with G M1 ganglioside and is the predominate A␤ species in diffuse plaques (2, 4 -7). Because of its prevalence in early appearing lesions, its lower solubility, and its tendency to form and stimulate the formation of fibrillar aggregates, it has been proposed that A␤42 serves to nucleate amyloid plaque formation (8 -11).
A model of this process appears to occur in solution, wherein a nucleation-dependent step in the process of fibril formation from A␤ proteins has been observed (8,10). This prompts one to look for conditions that promote nucleation by serving as "surrogate nuclei" or "folding templates." Strategies to design synthetic templates for ␤ sheet formation have been explored (12,13). Such templates need not be large, inasmuch as some rather small ligands are known to promote fibril formation (14).
The possibility that lipid membranes might serve as templates or nucleation sites for A␤ proteins was suggested nearly a decade ago (8), and an abundance of circumstantial evidence has appeared in the interim suggesting that lipid membranes may play an important role in the pathogenesis of AD. For example, the A␤ proteins that aggregate into fibrils, toxic or otherwise, are derived from amyloid precursor protein (APP), a membrane-anchored protein. Following its cleavage, A␤ proteins are associated with detergent-resistant lipid membrane domains in the brain (15), specific lipid components (5), or even its membrane-anchored parent, APP (16). Lipid membranes of varying composition strongly influence the rate at which A␤ proteins form fibrils (17)(18)(19)(20). Ultrastructural studies suggest that fibril formation tends to occur first in portions of diffuse deposits that are closest to membranes (21)(22)(23), and that intermediate forms give rise to pore-like assemblies within membranes (24).
We have previously observed that oxidatively damaged lipid membranes are much more effective than ordinary membranes at adsorbing A␤ proteins, and at misfolding them into pathological ␤ structure (25). This observation brings to mind the associations previously noted between oxidative stress, metal ions, and the toxicity of A␤ or APP (26,27), and indirect measures of lipid peroxidation that have more recently showed that oxidative stress correlates with amyloid plaque formation in animal models (28), and with disease severity in humans (29). It is also noteworthy that vitamin E, a lipophilic antioxidant, is relatively deficient in the post-mortem brain tissue of patients of Alzheimer's disease (30), and its administration in pharmacological doses appears to slow the clinical progression of disease (31).
We have now employed attenuated total internal reflectance-Fourier transform infrared (ATIR-FTIR) spectroscopy to inves-tigate further into the role of membranes and oxidative damage in the pathological misfolding of A␤ proteins. ATIR-FTIR is a powerful and increasingly popular technique for studying proteins, lipid membranes, and protein-lipid interactions (32)(33)(34)(35)(36). It is an especially powerful technique for the study of polypeptides that form ␤ structure because they exhibit a distinctive "splitting" of their amide IЈ absorption. It may not be clear whether this ␤ structure is parallel or antiparallel in mature amyloid fibrils (37), however, this question does not bear on the present analysis. ATIR-FTIR spectra of membrane-bound proteins are collected under conditions in which membrane lipids are highly ordered (as in ordinary lipid bilayers), and in which variables such as lipid composition, buffer composition, surface pressure, and temperature may be controlled. It yields quantitative information about the concentration, conformation, and orientation of various chemical groups on microgram samples. This degree of sensitivity enables the collection of complete spectra several times per minute for the assessment of adsorption kinetics.
These studies reveal much earlier events in fibrillogenesis than do measures of fibril formation based on dye binding, and demonstrate that oxidatively damaged lipid membranes and A␤42 are potently synergistic in promoting fibril formation by A␤40.
Lyophilized A␤40 and A␤42 were purchased from Bachem Bioscience Inc. (King of Prussia, PA) and American Peptide Co. (Sunnyvale, CA). Usually the materials from either source migrated as single bands in electrophoresis gels, and had expected molecular weights of 4330 and 4514 by mass spectrometry. However, some lots of material we received did not pass these tests. The proteins were dissolved in hexafluoro-2propanol and stored at Ϫ20°C. Before each experiment, aliquots of these stock solutions were evaporated, redissolved at a concentration of 90 M protein in 30 mM HEPES in D 2 O, pD 7.5, and ultrafiltered with 10-kDa cutoff filters immediately before use to remove preformed fibrils. Protein concentrations were determined with bicinchoninic acid reagent (Pierce).
Phospholipid Oxidation-Aqueous suspensions of pure SAPC in 50 mM Tris buffer, pH 7.5, were extruded at 25°C through 100-nm polycarbonate filters to produce unilamellar vesicles. A 100 M suspension of these vesicles was oxidized with H 2 O 2 (2 mM) and CuSO 4 (200 M) in 50 mM Tris buffer, pH 7.5, that had been presaturated with nitrogen gas. Peroxidation was monitored by following conjugated diene formation with UV absorption spectrometry at 234 nm (39). Oxidation was terminated by withdrawing aliquots at the prescribed times, adding 75 M BHT and 75 mM EDTA, followed immediately by vigorous extraction with two volumes of 2:1 (v/v) chloroform:methanol and storage at Ϫ80°C under argon until used. Phospholipid concentrations were determined by phosphate assay (38,40). Ultraviolet and mass spectrometric characterization of lipids oxidized in this manner have been published elsewhere (41).
Infrared Spectroscopy of Monolayers-Lipid monolayers were prepared in a Langmuir trough by applying lipids dissolved in hexane: ethanol to the surface of a buffer containing 30 mM HEPES in D 2 O at pD 7.5. The lipids applied in this manner consisted of 100% DMPC, a mixture of 20 mol % SAPC (unoxidized), 80 mol % DMPC, and 1 mol % BHT; or a mixture of 20 mol % SAPC (oxidized), 80 mol % DMPC, and 1 mol % BHT. The monolayer formed in the trough at the air-water interface was compressed to a surface pressure of 20 dynes/cm and applied onto a silane-treated germanium crystal as previously described (25,34,(42)(43)(44)(45)(46)(47). The protein was introduced in 8-g aliquots to a continuously stirred subphase compartment containing 6 ml of buffer.
ATIR-FTIR spectra were collected in rapid-scanning mode as 1024 co-added interferograms using a Bio-Rad FTS-60A spectrometer, a liquid-nitrogen cooled MCT detector, a resolution of 2 cm Ϫ1 , scanning speed of 20 MHz, triangular apodization, and one level of zero filling. An enclosure around the Langmuir trough is filled with argon to avoid spontaneous air oxidation of lipids at the air-water interface. As is typical for this instrument, non-level baseline correction, water vapor subtraction, and smoothing manipulations were not necessary and spectra are reproduced "raw" with only flat and level baseline correction. Spectra were fitted using IRfit, a procedure that fits a limited set of component bands simultaneously to several spectra (45). The aim of this analysis is to describe a set of related spectra with a minimum number of adjustable parameters. For the collection of spectra in kinetics mode, the resolution was set to 4 cm Ϫ1 , the scanning speed was increased to 40 kHz, and a complete spectrum was recorded every 15 s by coadding 78 interferograms. All spectroscopic studies were performed at 27°C, i.e. slightly above the phase transition temperature of pure DMPC. Vesicle Extrusion and Fibril Assay-Lipid mixtures consisting of 90% DMPC and 10% of either pure SAPC or oxidized SAPC in chloroform were evaporated, suspended in 30 mM HEPES, pH 7.4, by bath sonication, and extruded at 25°C through 100-nm polycarbonate filters to produce unilamellar vesicles. Aliquots of stock A␤ protein solutions in hexafluoro-2-propanol were evaporated and redissolved in phosphatebuffered saline (10 mM phosphate, 150 mM NaCl, pH 7.4) to yield concentrations of 0.16 mg/ml for A␤42 and 0.5 mg/ml for A␤40. The final concentrations of lipid, A␤42, and A␤40 in the aggregation mixture were adjusted to 10 M, 12 g/ml, and 48 g/ml, respectively, and incubated at room temperature with continuous gentle agitation for multiday intervals during which aggregation occurred. 30-l aliquots of these mixtures were added to 110 l of 10 M Congo Red in phosphatebuffered saline and incubated at room temperature for 30 min. Fibril formation was assayed by spectrophotometric measurements of Congo Red binding (48,49).
Negative Staining Electron Microscopy-Fibrillized A␤40 was adsorbed onto 300-mesh carbon-coated copper grids, stained with 1% aqueous uranyl acetate, and visualized with a JOEL 100CX transmission electron microscope (EM) (Peabody, MA). EM images were captured with a Hamamatsu digital camera (Bridgewater, MA) using AMT software (Danvers, MA).

Individual Proteins on Lipid Monolayer
Membranes-Monolayer membranes of three types were prepared for examination with ATIR-FTIR: saturated (100% DMPC), unoxidized (20% SAPC and 80% DMPC), and oxidized (20% oxidized SAPC and 80% DMPC). The monolayers were formed in a Langmuir trough and applied onto germanium crystals as previously described (46). 8-g injections of A␤ proteins were made into 6 ml of subphase buffer under the crystal yielding subphase concentrations of ϳ0.3 M. Protein accumulation was measured by calculating the total amide IЈ band area in spectra collected over 15-s intervals (a prime mark is added to amide I references when exchangeable hydrogens have been replaced with deuterium).
A␤40 accumulates to a significantly greater degree on oxidized membranes than on saturated or unoxidized membranes, similar in some respects to the behavior reported previously for A␤42 (25). However, the accumulation of A␤40 on oxidized membranes differed from that of A␤42 in two significant ways. First, A␤40 accumulation reached a plateau within 15 min and remained at this level for the remainder of the initial 60-min "seeding" period. In contrast, A␤42 accumulation on the oxidized membrane continued over the initial 60-min seeding period without reaching a plateau (Fig. 1). Second, the ATIR-FTIR spectrum of membrane-adsorbed A␤40 exhibited a prominent band at 1628 cm Ϫ1 but minimal absorption between 1680 and 1690 cm Ϫ1 . In contrast, the spectrum of membrane-adsorbed A␤42 exhibited the characteristic split amide IЈ band (Fig. 2, a and b).
Mixed Proteins on Oxidized Lipid Monolayer Membranes-After oxidized membranes were exposed to subphase buffers containing either 8 g of A␤40 or 8 g of A␤42 for 1 h, a series of injections containing 8 g of A␤40 were made into the subphase. In preparations initially exposed only to A␤40, subsequent injections of A␤40 up to a total of 40 g had little effect on the total amount of protein that accumulated (Fig. 1), or on the amide IЈ band shape. Similarly, injections of A␤42 under oxidized membranes initially seeded with A␤40 did not result in significant protein accumulation. In preparations initially seeded with A␤42, however, injections of A␤40 resulted in markedly increased protein accumulations. Moreover, each injection of A␤40 raised the integrated absorbance to a new plateau, indicating that membrane binding sites were not saturated, and suggesting that each plateau represented depletion of protein from the subphase by adsorption to the supported membrane. Protein could not be detected in the subphase with bicinchoninic acid reagent at these times, however, the starting concentrations of protein were only barely within detection limits.
Control experiments with saturated and unoxidized membranes seeded with A␤42 yielded results that were indistinguishable from those of the unseeded preparations (data not shown).
Other control experiments demonstrate that A␤40 does not effectively seed A␤42 accumulation, suggesting that the A␤40 that does accumulate interferes with the ability of A␤42 to promote protein accumulation. It should be noted that 1 mol % BHT was present in both the unoxidized and oxidized lipid samples and thus, cannot account for the behavior of A␤40 on oxidatively damaged membranes. As reported previously (25), BHT is added to all samples of unsaturated lipid because the effects of oxidative damage disappear after prolonged exposure of unsaturated monolayers to atmospheric oxygen. We conclude from these observations that oxidative damage and A␤42 work synergistically to cause the accumulation and misfolding of A␤40.
The data in our previous report describing the interaction of A␤42 with oxidatively damaged lipid membranes spanned only 60 min, and could not be extended because of environmental and instrumentation instabilities. The results in this report span continuous 4-h intervals. These much longer observation times were made possible by technical improvements that enhance signal stability and reduce noise, including tightly controlled ambient temperatures in the lab environment, higher output purge gas generators, and an enclosure around the instrumentation that helps stabilize temperatures and exclude water vapor by retaining exhausted purge gas around the outside of the spectrometer.
Mixed Proteins on Oxidized Lipid Vesicles-To determine whether the synergy between oxidatively damaged monolayer membranes and A␤42 could be demonstrated in bilayer membranes, and by more established means, we performed analogous experiments with unilamellar lipid vesicles using a Congo Red binding assay to detect fibril formation. The results shown in Fig. 3 compare unoxidized and oxidized lipid vesicles that were treated either with A␤40 alone or pre-treated with A␤42 followed by A␤40.
When 10 M A␤40 was mixed with 10 M unoxidized lipid in the form of vesicles, there was a marginally detectable amount of Congo Red binding to fibrils after 3 days. There was no significant binding detected when vesicles were composed of oxidized lipid, or when vesicles were not present (data not shown). With unoxidized lipid, pretreatment of lipid vesicles with 2.5 M A␤42 for 1 day before adding 10 M A␤40 also caused only a marginally significant increase in Congo Red binding after 2 days. We conclude that the conditions of these experiments give rise to minimal, if any, spontaneous fibrillogenesis over a 3-day incubation. These results are consistent with prior reports indicating that the concentration of A␤40 in our vesicle experiments is only slightly above its solubility limit, whereas the concentration of A␤42 is well below its solubility limit (9).
In contrast, the use of oxidized lipid and pretreatment with A␤42 FIG. 1. Integrated absorbance of the amide I band from 1600 to 1700 cm ؊1 over time following exposure of an oxidatively damaged lipid membrane to A␤ proteins. The membrane was supported by a 10 ϫ 50-mm germanium internal reflection crystal applied flat onto the membrane that had been formed at the air-water interface of a Langmuir trough. Proteins in the amounts indicated were injected through the membrane into 6 ml of buffer under the crystal in the trough subphase. At time ϭ 0, 8 g of A␤40 (solid line) or A␤42 (dots) was injected into the subphase. 8-g aliquots of A␤40 were also injected into the subphase at times ϭ 1, 2, and 3 h. Although the membrane seeded with A␤40 absorbed more protein in the first 15 min, it reached a plateau at this point and additional A␤40 did not yield large increases beyond this initial plateau. By 1 h, the membrane seeded with A␤42 had absorbed slightly more protein, and subsequent additions of A␤40 produced dramatic rises in protein adsorption. Results shown are the minute-by-minute average of three experiments conducted on different days. Results for A␤40 and A␤42 on unoxidized membranes (not shown) are indistinguishable from the A␤40-seeded membrane in this figure. in this type of experiment caused a marked increase in Congo Red binding after 2 days of incubation (Fig. 3). Examination of this preparation by electron microscopy after 3 days incubation showed that fibrils, 10 nm in diameter and 500-1000 nm long, had formed (Fig. 4).
Vesicles used in these experiments contained 10% unoxidized or oxidized lipid, whereas the monolayer experiments were performed with 20% unoxidized or oxidized lipid. The use of 20% oxidized SAPC was problematic in vesicle experiments because it enhanced fibril formation to such a degree that A␤40 fibrils formed after 3 days with or without seeding by A␤42. This suggests that oxidized lipids can promote A␤40 misfolding in the absence of A␤42. However, this observation was made at an A␤40 concentration of 10 M, i.e. slightly above its thermodynamic solubility limit in pure solution of 9 M (9). It was not detected in our ATIR-FTIR experiments where the A␤40 concentration was only 0.3 M in the subphase buffer. Reducing the unoxidized and oxidized lipid concentrations to 10% virtually eliminated fibril formation by A␤40 alone over the 3-day period of observation. These observations suggest that fibril formation by A␤40 is promoted by synergistic effects of oxidative membrane damage and pre-treatment with A␤42, and that the effect of oxidized membranes is dose-dependent.
Amide IЈ Band Shape Analysis-Amide IЈ bands from the infrared spectra of A␤40 and A␤42 are illustrated in Fig. 2. The amide IЈ band shape of A␤42 adsorbed onto oxidatively damaged lipid membranes from D 2 O is dominated by a low frequency component at 1622 cm Ϫ1 , but a smaller high frequency component is evident at 1684.0 cm Ϫ1 as well as a broad absorption centered between 1640 and 1670 cm Ϫ1 (Fig. 2a). This band shape closely resembles that of dry fibrillized A␤ protein segments (from H 2 O) previously reported (9).
The amide IЈ band shape of unseeded A␤40 on SAPC OX is also dominated by a low frequency component at 1629.1 cm Ϫ1 , and it has a broad absorption between 1640 and 1670 cm Ϫ1 , but the high frequency component at 1685.9 is only barely detectable (Fig. 2b). The amide IЈ band shape of A␤42-seeded A␤40 is likewise dominated by a low frequency component at 1627.8 cm Ϫ1 (Fig. 2c), but it differs from that of unseeded A␤40 in that there is substantially more absorption between 1680 and 1630 cm Ϫ1 and the high frequency component at 1685.9 cm Ϫ1 is more prominent. The spectra in Fig. 2, b and c, correspond to time ϭ 4 h in Fig. 1.
A 100 M solution of A␤40 at pH 7.4 was allowed to fibrillize for 10 days incubation at room temperature, after which time there is no detectable protein by assay in the supernatant of a centrifuged sample. This solution was then applied to an internal reflection crystal and the solvent was allowed to evaporate. The ostensibly "mature" fibrils in this preparation exhibit a prominent amide IЈ component at 1622.7 cm Ϫ1 , and a clear high frequency component at 1685.9 (Fig. 2d). These features are similar to those of A␤42 on SAPC OX (Fig. 2a). The amide IЈ band shape obtained from unfibrillized material (examined prior to the 10-day incubation) is dominated by a component at 1656.8 cm Ϫ1 , in addition to some minor components (Fig. 2e). It should be noted that the 1656.8 cm Ϫ1 component of this unfibrillized A␤40 cannot be detected in fibrillized A␤40 or in A␤ proteins that had adsorbed onto oxidatively damaged membranes. Thus, the 1656.8 cm Ϫ1 component appears to be characteristic of dry unfibrillized A␤40.
Quantitative analyses of these spectra were performed using IRfit (45), an approach designed to describe a set of related spectra under varying conditions with a minimum of fitting parameters. This approach is advantageous because it tends to identify components common to each spectrum with the same frequency, width, and shape. The results of this analysis are The experimental data were derived from 1024 co-added interferograms recorded at 2 cm Ϫ1 resolution, and reproduced without smoothing, water vapor subtraction, deconvolution, or other manipulations to enhance resolution or apparent signal to noise ratio. A flat level baseline correction has been applied. a, A␤42 absorbed to an monolayer containing oxidized lipids (SAPC OX ). b, spectrum of unseeded A␤40 on SAPC OX corresponding to time ϭ 4 h in Fig.  1. The absorption maximum is 0.0011. c, A␤42-seeded A␤40 on SAPC OX at time ϭ 4 h in Fig. 1. The absorption maximum is 0.0032, and the integrated area of the amide IЈ band is about 5-fold larger that that shown in panel b. d, dry A␤40 fibrils, formed after 10 days incubation in dilute HEPES buffer and evaporated onto an internal reflection crystal. e, dry A␤40 in dilute HEPES buffer immediately evaporated onto an internal reflection crystal. listed in Table I, and several observations should be noted. First, the dominant low frequency components in each of the fibrillized A␤ proteins (Fig. 2, a-d) all lie between 1620 and 1630 cm Ϫ1 . The amplitude of this component is 3-fold greater when the membrane is seeded with A␤42 (compare Fig. 2, b  and c). Because each component has the same width and shape, these amplitudes are proportional to their area. Second, each of the fibrillized A␤ proteins has two components at 1644.1 and 1669.7 cm Ϫ1 with the same position, width, and shape, but none bear evidence of the component at 1656.8 cm Ϫ1 that is characteristic of unfibrillized protein. Third, the spectrum from unfibrillized A␤40 requires a fitting component at 1656.8 cm Ϫ1 and it cannot be fitted with the components used to fit fibrillized A␤ proteins. Attempts to include a 1656.8 cm Ϫ1 component when fitting the spectra of fibrillized proteins yielded a component with zero amplitude, and this component could not supplant the components at 1644.1 or 1669.7 cm Ϫ1 . Interpretation of these amide IЈ band shapes is offered below.

FIG. 3. Congo Red binding over time to A␤ proteins incubated with lipid vesicles.
Each data point represents an average and standard deviation of three to five experiments. In experiments labeled "A␤42," this protein was incubated with the vesicles for 1 day prior to the addition of A␤40 on day 0. Because Congo Red binding is determined from the ratio of absorbance measured at two different wavelengths, the initial value is not necessarily zero, and the data is not normalized to zero. Electron micrographs of the samples containing A␤40, A␤42, and oxidized lipid vesicles at day 3 demonstrated classic 10-nm fibrils, 500 -1000 nm long.
FIG. 4. Electron micrographs of A␤40 fibrils negatively stained with uranyl acetate formed after a 3-day incubation with oxidatively damaged lipid vesicles that had been pretreated with A␤42 ("oxidized/A␤40/A␤42" sample at time ‫؍‬ 3 days in Fig. 3). The panel on the left includes a collapsed vesicle.

DISCUSSION
These results demonstrate that oxidative lipid damage is a promoter of A␤40 misfolding, aggregation, and fibril formation when pretreated with A␤42. Although seeding of A␤40 aggregation by A␤42 was observed a decade ago (9), the experiments described herein were performed below the concentration threshold where this occurs, and they constitute the first report that oxidatively damaged membranes work synergistically with A␤42 to promote A␤40 aggregation. This is a significant step forward in the development of an in vitro model of amyloid plaque formation because it provides an experimental system in which mechanistic details about the possible pathogenic relationships between oxidative stress, metal ions, A␤40/A␤42 deposition patterns, and A␤/APP toxicity may be explored (26 -29).
It is also significant that this link between oxidative damage and pathological misfolding of A␤40 was demonstrated using two independent techniques. Whereas it was possible to demonstrate this link using lipid vesicles and a Congo Red binding assay, it was much more quickly and clearly detected using supported lipid membranes and a specialized form of ATIR-FTIR spectroscopy. The latter approach offers at least three significant advantages over dye binding assays. First, speed, reporting on events that occur over minutes rather than days. Second, sensitivity, with clear results from 0.3 M solutions of each protein, compared with 2.5 M A␤42 and 10 M A␤40 for dye binding. Assays using thioflavin T fluorescence (50) are more sensitive than Congo Red, but still not nearly as sensitive as ATIR-FTIR and they do not offer a speed advantage. Third, additional information, about adsorption kinetics, secondary structure, and the organizational state of membranes lipids. The vesicle experiments confirm the ATIR-FTIR results, and are key control experiments showing that negative results in our ATIR-FTIR experiments are not merely because of proteins leaving the membrane surface and becoming undetectable by ATIR-FTIR.
Jarret et al. (9) demonstrated the potential of A␤42 (below its solubility limit) to seed fibril formation by A␤40 (above its solubility limit) over the course of several days. The current data expand on these much earlier results by reporting the existence of an interaction within minutes of mixing, by using concentrations of both proteins well below their solubility limits, and by demonstrating that oxidative damage works synergistically with A␤42 to promote A␤40 misfolding.
Insofar as the internal reflection technique reports on adsorption and folding events that occur within minutes of exposing proteins to oxidatively damaged membranes, we do not know whether the proteins being detected have formed short segments of true fibrils, protofibrillar intermediates, interme-diates with helical structure (51), or another as yet undescribed stage of fibril formation. In any case, it is clear that the results from internal reflection predict and correlate to the formation of classic amyloid fibrils that bind Congo Red and can be visualized by electron microscopy. At this point, technical obstacles preclude the examination of membrane-adsorbed A␤40 by electron microscopy or atomic force microscopy.
It has been suggested that the A␤42 used to seed membranes in these experiments may have formed small seed fibrils even before it encountered the membrane. We cannot rule out this possibility, but the proteins are treated with hexafluoro-2-propanol immediately prior to use, and after redissolving the proteins in buffer, they are ultrafiltered. This reportedly yields a structurally homogenous non-fibrillar protein preparation (52). In any case, and irrespective of its aggregation state, A␤42 seeding was only effective at promoting A␤40 misfolding on oxidatively damaged membranes: A␤42 did not adsorb or misfold on undamaged membranes, and it did not promote this behavior by A␤40 on undamaged membranes. Thus, the presence of seed fibrils in our A␤42 preparations would not diminish the significance of oxidative damage to the process of fibril formation.
The interpretation of amide IЈ band shapes in terms of conformational assignment is fraught with pitfalls, especially when there are differences in sample preparation. One must be circumspect when comparing the spectra of membrane-adsorbed spectra (Fig. 2, a-c) to those of dried protein (Fig. 2, d  and e). For example, the 1656.8 cm Ϫ1 component of dry unfibrillized A␤40 (Fig. 2e) does not correlate well to a secondary structure category. One would expect the protein to be random coil under these conditions (53), but random coil polypeptides only absorb at this frequency in H 2 O. All exchangeable hydrogen atoms in the protein samples subjected to infrared analysis in this work have been thoroughly replaced with deuterium, as evidenced by the lack of any residual amide II band between 1600 and 1500 cm Ϫ1 . Under these conditions, the characteristic absorption band of a random coil protein shifts to 1643 cm Ϫ1 (54). The value of this spectrum is not its interpretation in terms of secondary structure, but its distinctiveness from that of Fig. 2, a-d; it demonstrates that unfibrillized A␤ protein cannot be detected in fibrillized A␤40 under the same conditions, or in A␤ proteins when they are membrane-adsorbed.
The value of the spectrum from dry fibrillized A␤40, on the other hand, is its similarity to the spectra from membraneadsorbed A␤ proteins, especially A␤42 and A␤42-seeded A␤40. Despite differences in preparation (adsorbed at the watermembrane interface versus dried), the spectra in Fig. 2, a,   at 1684 -1686 cm Ϫ1 . This pair of components has been experimentally and theoretically correlated with extended multiplestranded ␤-sheet structures formed in other proteins (55,56). Unseeded A␤40 is distinguished from A␤42-seeded A␤40 by having only a trace of the high frequency component, and an overall absorption intensity that is only one-third as large after 4 h of accumulation. Nonetheless, A␤42-seeded A␤40 is distinct from mature A␤40 fibrils by the blue-shift of its low frequency to 1627.8 cm Ϫ1 and the prominence of a mid-amide IЈ component at 1644.1 cm Ϫ1 . This suggests that the A␤40 that has adsorbed onto an oxidatively damaged membrane is immature with respect to folding, and that after 4 h it has still not fully adopted the conformation it will eventually adopt in a mature fibril. The amide IЈ components at 1644.1 cm Ϫ1 cannot be interpreted definitively. This is a relatively broad component that may well have irresolvable subcomponents. Furthermore, secondary structure assignments for amide IЈ components between 1680 and 1630 cm Ϫ1 are imprecise. Bands at 1648 -1655 cm Ϫ1 are generally associated with ␣-helical conformation, whereas bands at 1643 cm Ϫ1 are traditionally associated with unordered structure (54,57). This interpretation would suggest that A␤42-seeded A␤40 adsorbs onto the membrane in a random configuration and that it subsequently adopts a less nonrandom fold as it matures into an amyloid fibril. More recent studies, however, have demonstrated amide IЈ bands in helical proteins far outside this range, even as low as 1632 cm Ϫ1 (58). Thus, one cannot rule out the possibility that a portion of the 1644.1 cm Ϫ1 component in A␤42-seeded A␤40 represents a helical structure, possibly even the oligomeric helical intermediate state detected in circular dichroism studies (51,53).
A preliminary characterization of the oxidatively damaged lipids used in this work has been published showing that the preparation contains hydroperoxides, conjugated dienes, and isoprostanes (41). This analysis also showed that the protocol employed in this work actually damages only a very small percentage of the SAPC used in forming supported membranes, or vesicles. It is difficult to quantify the degree of damage with precision because most of the reactant lipid remains undamaged (ϳ98%) while the products of damage are legion. Moreover, chemical characterization of the products of damage is a daunting task, essentially requiring "lipidomics" technologies that have yet to be developed. Thus, further detail about the chemical nature of the active lipid species in these experiments is not available at this time.
We have observed that oxidatively damaged lipids promote A␤42 misfolding when monolayers are compressed to 20 mN/m before their application to the germanium internal reflection crystal support, but this does not occur if they are compressed to 36 mN/m. 2 Because the higher monolayer pressure corresponds more closely to widely accepted values for the bilayerequivalent lateral pressure of a monolayer (59), one possible interpretation of these observations is that oxidized lipids may not promote A␤ accumulation and misfolding in the bilayer membranes of cells. However, our experiments were not conducted on monolayers at an air-water interface, but rather on monolayers supported by a firm hydrophobic alkyl-silane surface. The pressure to which a monolayer must be compressed to yield a bilayer-equivalent lateral pressure when supported on such a surface is not clear. This uncertainty underscores the importance of our experiments performed with bilayer vesicles: they corroborate the monolayer experiments performed at 20 mN/m by demonstrating that oxidized lipids in bilayers promote true amyloid fibril formation, and by establishing that the ability to synergize with A␤42 in promoting A␤40 aggregation is a bilayer-equivalent property of supported monolayers.
We have demonstrated the synergistic effects of oxidative damage and A␤42 in vitro on synthetic membranes, whereas membranes in the brain are vastly more complex mixtures. Therefore, one must keep in mind that it is the more complex membrane, the one containing myriad oxidation products, that accelerates the accumulation and folding of A␤ in these experiments. We have verified the presence of the isoprostane 8,12iso-iPF 2␣ -VI in our lipid preparations (41), and the other 31 isoprostanes are likely to be present as well. In addition, we can safely assume that a diverse assortment of conjugated diene species, oxygen-addition products, and various partially degraded lipid forms will also be present. Hence, the membranes promoting amyloidogenesis in this work are complex and heterogeneous. It is the simple and homogeneous membranes that do not accelerate the accumulation and misfolding of A␤.
It is not yet known how oxidative damaged membranes promote the misfolding of A␤ proteins, but one can readily imagine them acting kinetically as "seeds" or thermodynamically as "templates." Either way, they become sites at which there is a locally high concentration of the pathological form. It is also not known at this point whether membranes, damaged or otherwise, promote the pathological folding of A␤ in vivo. In pursuing answers to these questions, one must keep in mind that seeds and templates would not need to persist for long periods of time, nor would they need to involve large areas of membrane, to promote pathological conformational changes. The transient formation of a small amyloidogenic patch of membrane may be sufficient to create a nucleus of misfolded A␤ proteins that promotes the misfolding of additional protein with no need to maintain the amyloidogenic patch.