The effect of cholesterol and monosialoganglioside (GM1) on the release and aggregation of amyloid beta-peptide from liposomes prepared from brain membrane-like lipids.

In order to investigate the influence of cholesterol (Ch) and monosialoganglioside (GM1) on the release and subsequent deposition/aggregation of amyloid beta peptide (Abeta)-(1-40) and Abeta-(1-42), we have examined Abeta peptide model membrane interactions by circular dichroism, turbidity measurements, and transmission electron microscopy (TEM). Model liposomes containing Abeta peptide and a lipid mixture composition similar to that found in the cerebral cortex membranes (CCM-lipid) have been prepared. In all, four Abeta-containing liposomes were investigated: CCM-lipid; liposomes with no GM1 (GM1-free lipid); those with no cholesterol (Ch-free lipid); liposomes with neither cholesterol nor GM1 (Ch-GM1-free lipid). In CCM liposomes, Abeta was rapidly released from membranes to form a well defined fibril structure. However, for the GM1-free lipid, Abeta was first released to yield a fibril structure about the membrane surface, then the membrane became disrupted resulting in the formation of small vesicles. In Ch-free lipid, a fibril structure with a phospholipid membrane-like shadow formed, but this differed from the well defined fibril structure seen for CCM-lipid. In Ch-GM1-free lipid, no fibril structure formed, possibly because of membrane solubilization by Abeta. The absence of fibril structure was noted at physiological extracellular pH (7.4) and also at liposomal/endosomal pH (5.5). Our results suggest a possible role for both Ch and GM1 in the membrane release of Abeta from brain lipid bilayers.

The pathology of Alzheimer's disease (AD) 1 includes extracellular amyloid plaques, intraneuronal neurofibrillary tangles, synaptic loss, and neuronal cell death. The major components of amyloid plaques are the amphiphilic 40 and 42 residue peptides, A␤-(1-40) and A␤-   (1,2). Amyloid ␤-peptide (A␤) consists of a hydrophilic N-terminal region (residues 1-28) and a hydrophobic C-terminal region (residues 29 -40 or 29 -42). The hydrophobic part of A␤ is originally part of a transmembrane ␣-helix of APP anchored in the membrane of several subcellular compartments, including the ER (3). Proteolysis by the enzyme(s) ␥-secretase leads to the formation of A␤ within the membrane. Thus, the membrane release of A␤ following this enzyme cleavage should play a pivotal role in subsequent amyloid plaque formation.
Recent studies have shown that the interaction of A␤ and lipids plays an important role in the pathogenesis of AD. For instance, the fibrillogenic properties of A␤ are in part a consequence of the composition of the membrane in which it resides, its peptide sequence, and its mode of assembly within the membrane (4). In terms of membrane composition, Ch and GM1 in neuronal cell membranes are widely accepted to be modulators of membrane-associated A␤ fibrillogenesis and neurotoxicity (5,6). The formation of GM1-bound A␤, which is thought to be a seed for the formation of toxic amyloid fiber, depends on the concentration of Ch in model membranes prepared from GM1/Ch/sphingomyelin (SM) (7). Additionally, oligomeric A␤ can promote the release of lipids from asterocytes and neurons by forming A␤-lipid particles consisting of Ch, phospholipids, and GM1 (8).
It has been suggested that A␤-  is essential to the early development of AD pathology but is not alone sufficient to promote the formation of mature neuritic plaques unless it is succeeded by the deposition of A␤-(1-40) (9). Compared with A␤- , A␤-  has been shown to have a greater potential for aggregation (10). Studies of A␤-lipid interaction using total brain lipid extract have shown that the peptides interact in different ways: 1) A␤-(1-40) destabilizes model membranes and 2) A␤-  initially destabilizes but then with time proceeds to stabilize the membrane (11). These findings are consistent with a "seeding" hypothesis, in which the aggregates of A␤-(1-42) act as an initiation factor for early plaque formation, which is then followed by the progressive accumulation of A␤-  in the AD brain. So too, they provide an insight into a mechanism of fibrillogenesis, which is at least in part controlled by the release of A␤ from the membrane.
AD is a disease that involves attack of the central cerebral cortex. To investigate how Ch and the ganglioside GM1 may influence the release of A␤-  and A␤-(1-42) from cerebral cortex membranes we have prepared and examined some model A␤-containing liposomes. The liposomes are composed of a lipid-mixture similar in composition to cerebral cortex membranes (CCM-lipid). The liposomes prepared were of four different lipid compositions (see Table I); CCM-lipid, liposomes with no GM1 (GM1-free lipid), liposomes without cholesterol (Ch-free lipid), and liposomes without GM1 and Ch (Ch-GM1-free lipid). Liposomes were prepared by hydrating an organic film composed of a mixture of A␤ and one of the four lipid mixtures outlined in Table I. This method of liposome preparation was chosen to reflect a more natural release process as opposed to a method involving addition of A␤ to preformed liposomes.
The release of A␤ occurs from plasma, endosomal, lysosomal, and Golgi membranes by proteolysis (12)(13)(14). All experiments were done at a pH of 7.4 to represent a physiological extracellular pH. However, to model whether or not the more acidic environment of the endosomal pathway is significant we have carried out a number of our experiments at a pH of 5.5. The common endosomal and lysosomal pH values are thought to be 6.0 and 4.5-5.5, respectively (15,16); consequently we chose 5.5 to be representative of both organelles. We report here the mode of interaction of A␤-(1-40) and a mixture of A␤-(1-40)/ A␤-(1-42) (10/1) with each of the four types of liposome as revealed by CD, turbidity, and negative-stained TEM measurements.
It has been recognized that the self-assembly of A␤ is dependent on the initial A␤ structure, i.e. whether it is monomeric and has a random coil structure or whether it has a ␤-sheet structure (17,18). In the present study, the lyophilized powders of the peptide(s) were dissolved in HFIP, a solvent well known for its good solubilizing and ␣-helical structure-promoting properties. In HFIP solution, A␤ is monomeric (19 -21). The peptide stock solution was made by dissolving the lyophilized amorphous powder in HFIP. Each liposome, composed of the appropriate mixture of lipids and peptide, was prepared by hydration and sonication of a film obtained from the evaporation of a mixture of the lipids in CHCl 3 solution and A␤ in HFIP solution.
CD Spectrum Measurements-Mixed films of lipids and A␤-(1-40) or A␤-(1-40)/A␤-(1-42) (10/1) in HFIP were prepared by evaporating the organic solvents under a stream of nitrogen. Any residual solvent in the film was removed in vacuo overnight. Films were hydrated in Tris buffer solution (5 mM Tris, 100 mM NaCl, pH of 7.4 or 5.5) and sonicated by ultrasonic irradiation in the cuphorn of a Branson Model 185 sonifier (Danbury, CT). Solutions were sonicated at room temperature for about 30 min until they became transparent. CD spectra were measured on a JASCO J-600 apparatus (JASCO, Tokyo, Japan) controlled by a personal computer (NEC PC-9801) using a 1-mm pathlength quartz cell at 25°C. Four scans were averaged for each sample. An averaged blank spectrum (vesicle suspension or solvent) was subtracted from each sample spectrum. The peptide and lipid concentrations were 50 M and 1 mM, respectively. Conformational analyses were performed using CONTIN3 in a software package for analyzing protein CD spectra (CDPro) via the Internet (22). In our experiments, the program has been successfully applied to the solutions of peptides consisting of ␣-helix, ␤-sheet structure, ␤-turn, and random structure such as the p53 tetramerization domains that have been well characterized by NMR (23).
Turbidity Measurements-A mixed solution of lipids in chloroform and A␤ in HFIP measured to the appropriate peptide/lipid mol ratio was placed in a round bottom flask and dried under a stream of N 2 gas. The residual films were further dried overnight in vacuo and then hydrated with Tris buffer saline solution (5 mM Tris, 100 mM NaCl, pH of 7.4 or 5.5) by vortexing and then sonicated as described above. Lipid concentrations were kept to a concentration of 100 M in the same Tris buffer at 25°C. The absorbance of the sample solution was recorded at 400 nm using a JASCO spectrometer (JASCO Corp., Tokyo, Japan) after vigorous vortexing.
Transmission Electron Microscopy (TEM)-Each sample was absorbed onto a carbon-coated copper grid (mesh) by floating a drop of sample solution. Excess solution was removed by filter paper blotting, the grid was washed by floating a drop of water, and then the water was removed. The sample on the grid was then negatively stained with an aqueous phosphotungstic acid (1.0%), and the excess staining solution was removed. After drying, the samples were imaged with a HITACHI HU-12A electron microscope (Hitachi, Japan) operating at 100 kV. Samples prepared for the CD experiments were also used for the TEM experiments. The peptide and lipid concentrations were 50 M and 1 mM, respectively.

Compositions of Lipid Mixtures Similar to Cerebral Cortex
Membrane-To investigate the roles of GM1 and Ch in A␤-lipid interactions, we chose to model cerebral cortex membranes, because this is the main location of amyloid in the body. Values for the composition of total lipid extract from cerebral cortex membranes were those reported by Rossiter (24). Because of its easily oxidative property and its instability under acidic conditions, plasmalogen was not included in the model liposomes of this study. In the human cerebral cortex tissue, there is about 10% (w/w) of plasmalogen. We have adjusted the proportion of the other components to allow for this omission as shown in Table I, where we list the compositions of the four types of liposomes examined. Ch (26 mol %) is significantly more abundant in CCM membranes than GM1 (4 mol %).
Secondary Structure of A␤-  in Lipid Bilayers-Amyloidogenesis involves a transition from random or ␣-helical to a ␤-structure, which is necessary for fibril formation in vivo. Thus we monitored the conformational change of A␤-(1-40) or A␤-(1-40)/-(1-42) (10:1, molar ratio) in the four liposome systems (Table I) by CD spectroscopy over a period of 7 days. In buffer solution, the peptide A␤-(1-40) adopts a mainly random coil structure upon addition of the peptide stock solution to the buffer solution (Fig. 1A). This result agrees with the CD data reported previously (7,25). Then, after 5 days, it takes a mainly ␤-sheet structure as suggested by the peak minimum around 218 nm. A conformational analysis of its spectrum showed 15% ␣-helix and 30% ␤-structure. This main structure still persisted after 7 days.
In the CCM membrane at pH 7.4 ( Fig. 1, B-a), upon preparation from the lipid-peptide film, A␤-(1-40) adopts a mainly random structure within the liposomes, however, after 1 day a

Effect of Cholesterol and GM1 on Release of A␤ from Liposomes
broad negative band is seen around 218 nm, the intensity of which is approximately equal to that of the peak observed for the 7-day-old sample in buffer. But conformational analysis of the spectrum showed 60% ␤-structure and the same spectral pattern after 3, 5, and 7 days. When compared with the CD spectra in buffer, the ␤-structure formation was accelerated and more extensive. At pH 5.5, similar CD curves were obtained, though the rate of ␤-structure formation was significantly reduced. After 1 day under acidic conditions there was only 36% ␤-structure as compared with 60% at pH 7.4 ( Fig. 1, On the other hand, the peptide in GM1-free lipid lipo-somes at pH 7.4 ( Fig. 1C), showed a shallow negative band around 210 nm with 25% ␣-helix and 10% ␤-structure, and after 1, 3, and 5 days the negative band had shifted to 216 nm, corresponding to 45% ␤-structure and 10% ␣-helix, indicating that the absence of ganglioside slightly reduces formation of ␤-structure, compared with CCM membranes. Interestingly, in Ch-free membrane (Fig. 1D), A␤-(1-40) had a negative band around 205 nm similar to that of GM1-free liposomes after preparation. Some ␣-helical structure (about 15%) was seen, but in time the band around 205 nm intensified, and a crossing point on the horizontal axis occurred characteristic of ␤-struc- ture. After 3 days, ␤-structural content reached 50%, indicating, that in time, the absence of either Ch or the absence of GM1 led to a decrease in ␤-structure formation. For Ch-and GM1-free liposomes (Fig. 1E) at pH 7.4, A␤-(1-40) showed a broad negative band around 220 nm upon preparation, there was some ␣-helical structure present. However, in time, the band shifted to 203 nm and became shallower, also the crossing point of the horizontal axis red-shifted to more than 210 nm, indicating decreased ␤-structure. In fact, the spectra could not be analyzed using the CDPro software (22). Similar CD spectra have been reported in which the spectrum has an atypical minimum at 223 nm and a maximum at 203 nm; this pattern is characteristic of amyloid fibers (26,27). However, electron microscopy of A␤-(1-40) did not detect any fibril structure in Ch-free or Ch-GM1-free membranes. These results indicated that the CD spectral patterns having the minimum at 223 nm and the maximum at 203 nm do not necessarily exhibit the characteristics of amyloid fibers. At pH 5.5, although the minima around 220 were shallower than those at pH 7.4, very similar CD curves were observed, and the fibril structure was not detected by TEM at either pH (data not shown). Consequently, it would appear that the presence of Ch and GM1 ganglioside is necessary for ␤-amyloid formation to occur.
CD Spectra of A␤-(1-40)/A␤-(1-42) (10:1)) in Lipid Bilayers-To investigate a seeding hypothesis in which aggregates of A␤-(1-42) act as an initiation factor for early plaque formation, we examined changes in conformation of A␤-(1-40)/-(1-42) (10:1, molar ratio) in the four model liposomes at pH 7.4 (Fig. 2) by circular dichroism spectroscopy. Our results show that in the CCM and Ch-free membranes, the rate of ␤-structure formation is greater than that for the corresponding A␤-  liposomes. In CCM-lipid, upon preparation, we observed 40% ␤-structure which, after 1 day became 50%; for A␤-  2C and 1D). Interestingly, in the Ch-GM1-free lipid liposomes upon preparation the peptide had a 35% ␤-structure and 10% ␣-helical structure. This was not observed for A␤-(1-40) and in time it transformed the conformations to the other conformational mixture containing ␤-structure as described above (Fig. 2D). These results indicate that the rate of formation of the ␤-structure is generally promoted by the addition of A␤-(1-40), although ultimately formation depends on lipid composition.
Turbidity Measurements-To monitor the aggregation or precipitation produced by lipid-peptide interactions, we measured the changes in turbidity in the buffer solution and solutions of the four A␤-lipid mixture liposomes over a 7-day period (Fig. 3). Solution absorbance measured at 400 nm was plotted as a function of time. Days after successive film preparation, hydration, and sonication of mixture solutions for different ratios of A␤-(1-40) and lipids are shown on the abscissa. The lipid concentration was kept at 100 M. In the absence of liposomes, at a peptide concentration of 40 M at pH 7.4, the absorbance gradually increased from 1 day and reached a maximum at 3 days, and then leveled out as a plateau (Fig. 3A). However, at lower peptide concentrations (5-20 M), the aggregation behavior was dependent on peptide concentration. In conjunction with the CD results, it was concluded that increasing turbidity was attributable to ␤-structure formation, associated with A␤ aggregation.
For 40 M A␤ in CCM liposomes at pH 7.4, solution turbidity gradually increased over 4 days, then it slightly decreased, but at lower peptide concentrations there was little change in turbidity (Fig. 3B). With the results of CD and TEM experiments in mind, little or no change in turbidity is due to an aggregation of A␤-(1-40) to form a ␤-structure. Similar turbidity changes in CCM were obtained at pH 5.5 (date not shown). In the GM1free liposomes at pH 7.4, solution turbidity did not change at the examined A␤ concentrations except for 40 M peptide, where a slight decrease occurred 1 day after preparation and was followed by a gradual increase (Fig. 3C). Little or no change in turbidity was consistent with the CD and TEM results, which showed the formation of aggregated peptidelipids particles (to be described below). Interestingly, in Ch-GM1-free lipid liposomes at pH 7.4, turbidities for 20 and 40 M peptide decreased. For 40 M peptide an especially dramatic decrease in turbidity occurred over the first day (Fig. 3, E-a). Similar changes in turbidity were observed at pH 5.5 (Fig. 3,

FIG. 2. CD spectra of A␤-(1-40)/A␤-(1-42)(10/1) in CCM (A), GM1-free (B), Ch-free (C), Ch-GM1-free (D) over a period of 7 days after the incubation of peptides to liposomes.
Shown are spectra immediately after preparation (black), at 1 day (green), 3 days (red), 5 days (orange), and 7 days (blue). Peptide and lipid concentrations are 50 M and 1 mM, respectively. E-b). The decrease in turbidity indicated liposomes were solubilized by A␤- , to form small peptide-lipid particles. The TEM image for this solution showed the liposomes became smaller with time and showed no fibril structure. Although a similar decrease in turbidity was observed for the Ch-free liposome solution (Fig. 3D), this was independent of peptide concentration; the turbidity of 20 M A␤ was much larger than that of 40 M. This also indicated that liposomes were solubilized and led to the formation of small particles.
Turbidity of A␤-(1-40)/A␤-(1-42) (10:1) was measured in CCM lipid and ChGM1-free lipid at pH 7.4 (data not shown). Initial turbidity did not depend on peptide concentration and did not change dramatically with time, which indicated the membrane solubilization did not occur as it did for A␤-(1-40) in Ch-free and Ch-GM1-free membranes.
Studies by Transmission Electron Microscopy-To observe changes with time in the morphological characteristics of A␤-(1-40) and A␤-(1-40)/A␤-(1-42) (10:1) both in buffer and in the liposomes we made measurements by TEM. Liposomes were prepared from a mixture of lipids (1 mM) and peptide (50 M) and then examined over a period of 1-10 days under the electron microscope. The TEM samples were negatively stained. For A␤-(1-40), after 1 day in buffer solution at pH 7.4, an indistinct fibril structure was observed, which after 10 days became better defined (Fig. 4A). In CCM lipid membrane at pH 7.4, a distinct fibril structure was observed, with a dark phospholipid membrane-like shadow after 1 day (Fig. 4B-a, i). After 10 days the same clear fibril structure seen in saline buffer was observed (Fig 4B-a, iii). At pH 5.5, lots of short filaments were visible after 1 day and clear long fibrils were observed after 3 days, indicating that the growth of fibril structure is slower than that at pH 7.4 (Fig 4B-b). This is consistent with the slower rate of ␤-structure formation seen by CD after 1 day at pH 5.5 relative to that at pH 7.4. In GM1-free membranes (Fig.  4C) at pH 7.4 after 1 day, long fibril structures were present along the surface of vesicles, which were probably large aggregated/fused liposomes. After 3 days, small vesicles emerged around the large aggregate liposomes, also some fibrils were still apparent along their surface. After 10 days we observed aggregates of small vesicles (several ten-fold nanometers in diameter). These results indicate that A␤ can disrupt the large liposomes into smaller vesicles, presumably by the formation of lipid-peptide complexes. After 1 day in the Ch-free membrane (Fig. 4D) at pH 7.4, a few relatively long fibrils were observed around the large liposomes. After 4 days, numerous small filament-like structures, probably lipid-peptide complexes, were present around the liposomes. Interestingly, after 10 days, thicker and longer fibrils (several ten-fold nanometers in diameter) were observed around the liposomes, which were different from the fibril structure observed in the buffer and CCM liposomes.
These results indicate that although the Ch-free membrane first releases some A␤ to create fibrils, in time, the membranes were slowly solubilized by A␤ to make short and thin fibrils and eventually thicker and longer fibers. In Ch-GM1-free membranes at pH 7.4 and 5.5 (Fig. 4, E-a and -b) fibrils were not observed, liposomes of various shapes and sizes were present, and with time an increase in the number of smaller vesicles occurred.

FIG. 4. Electron micrographs in buffer solution (A), CCM at (a) pH 7.4 and at (b) pH 5.5 (B), GM1-free (C), Ch-free (D), Ch-GM1-free (E) at (a) pH 7.4 and at (b) pH 5.5 over a period of 14 days after the incubation of A␤-(1-40) and in CCM (F) and Ch-GM1-free liposomes (G) of A␤-(1-40)/A␤-(1-42)(10:1)
. Images: 1 day (i), 3 days (ii), 10 days (iii) after liposomes were prepared from 50 M peptide and 1 mM lipid solutions, respectively. Bars are 500 nm. free membranes, both fibrils and spherical liposomes were present after 1 day, although with time the number of fibrils increased and the spherical liposomes disappeared (Fig 4G). Similar images (data not shown) were observed for the Ch-free membranes. DISCUSSION Because A␤ peptides are generated by the partial processing of the transmembrane ␣-helix of APP anchored in the brain membrane, their release from the membrane must play an important role in their subsequent aggregation and precipitation. Thus, to investigate how membrane lipids participate in the formation of fibril structure, we monitored the release of A␤ peptide from A␤ peptide-model membranes. Liposomes were prepared from A␤ and a lipid-mixture similar in composition to that of cerebral cortex membranes with and without ganglioside and/or cholesterol.
Our CD studies have shown that when a 50 M solution of A␤-  in HFIP solution is hydrated in buffer solution, it undergoes a transition from random coil to ␤-structure over a period of 3 days (Fig. 1A). This is consistent with the turbidity measurements. Over 24 h, the turbidity did not change, but it rapidly increased from 1 to 3 days, and only moderately after 3 days (Fig. 3A). From these results we propose that hydration of monomeric A␤-(1-40) in organic solvent caused a slow change in conformation from random coil to ␤-structure. This ␤-struc- ture then serves as a seed in the rapid formation of an extensive ␤-sheet structure (28). TEM measurements supported the presence of an extended fibril formation just 3 days after peptide hydration.
In CCM membranes at physiological pH (7.4), A␤-(1-40) was mainly random coil after preparation, although some ␣-helical and ␤-sheet content was present (Fig. 1B). After 1 day, mainly ␤-structure had formed which persisted for 10 days. The turbidity increased moderately only for high concentrations of A␤-(1-40) (Fig. 3B-a). TEM images of fibrals seen after 1 day (Fig. 4B-a, i) were clearer than those observed for the buffer solution. At pH 5.5 ( Fig. 4B-b, i) the presence of fibrils after 1 day is uncertain, although after 3 days an extensive fibril formation is apparent (Fig. 4B-b, ii). For the CCM liposomes at endosomal pH the rate of formation of ␤-structure and fibrils is definitely slower than at physiological pH; however, the final resulting fibril structures appear the same. That A␤ fibril formation is pH-dependent has been reported in the literature (43).
From the CCM liposomes A␤-(1-40) was released rapidly, and resulted in the formation of a fibril structure. In contrast, in GM1-free liposomes, A␤-(1-40) consisted of a mixture of ␣-helix and ␤-structure and with time the proportion of ␤-structure increased. However, a slight increase in turbidity was observed for 40 M A␤. Its TEM image showed at first an incomplete short fibril structure around the lipid surface, and in time, small vesicles began to emerge, until finally only aggregates of small vesicles were visible (Fig. 4C). Apparently, A␤ was able to solubilize the ganglioside-deficient membrane into small vesicles.
In Ch-free Lipid liposomes, a similar change from ␣-helix to ␤-structure was observed. The turbidity decreased drastically for 20 M A␤, but not so for 40 M. However, there were few signs of fibril structure after 1 day; instead spherical liposomes were visible (Fig. 4D, i). From 3 to 14 days, a gradual thickening and elongation of the thin, short fibrils around the spherical vesicles took place (Fig. 4, D, ii and iii). These fibrils were different in appearance to those observed in the buffer and CCM membranes. We propose these were peptide-phospholipid membrane complexes, because we have shown previously that highly hydrophobic peptides can form nanotubular fiber structures (29 -31). Ch may assist in fibril formation, by promoting the release of A␤-(1-40) from natural membranes. Interestingly, in both GM1-and Ch-free liposomes, the CD spectra show a shallow minimum around 223 nm (Fig. 1D), which was not seen for the CCM and GM1-free liposomes. Moreover, the TEM showed liposomes of various shapes and sizes, but no fibril structure was seen. These phenomena were the same at neutral and acidic pH membrane solutions. This suggests that the coexistence of Ch and ganglioside in the CCM membrane has a crucial role in the release of A␤-(1-40) from the membrane and the subsequent formation of fibril structures.
A recent study has shown that in different lipid membranes A␤ can follow two pathways of assembly: pathway 1) the formation of fibril structure in the presence of acidic lipids; and pathway 2) the formation of small aggregates (but no fibril structures) in the presence of neutral lipids (32).
This study shows GM1-free membranes promote the formation of ␤-structure and small peptide-lipid vesicles, several ten-fold Angstrom in diameter. This process may take pathway 2; the absence of GM1 leads to the decrease in the acidity of the membranes, resulting in the elimination of A␤-fibril formation. However, membrane disruption still occurs via the expansion of aggregated A␤ through the bilayers, resulting in the solubilization of membranes to form small peptide-lipid vesicles. Matsuzaki and Horikiri (33) reported that A␤ has a high affin-ity for GM1 ganglioside in the bilayer and is able to form a ␤-sheet structure. It has been reported that the tight binding of A␤ is to the sialic acid group of the GM1 (34). In the presence of GM1, A␤ probably follows pathway 1, a conformational transition from ␣-helix to a ␤-structure leading to fibril formation. However, tight binding of the peptide to GM1 may prevent its release from the membrane, resulting in accumulation of peptide and formation of a ␤-sheet scaffold structure. This may be the critical nucleus for fibril formation. After nucleation, fibril growth through the lipid bilayer results in destabilization of the membrane, leading to amyloid deposition or the formation of lipid particles (to be described below).
Ch promotes fibril structure formation as we observed by TEM; in the absence of Ch, a well defined fibril structure was not visible even after a few days. Interestingly, in the absence of Ch or GM1 A␤ does form a ␤-structure, so the formation of a ␤-structure does not necessarily lead to fibril formation. An increase in Ch in the membrane results in increased membrane stiffness and a decrease in membrane fluidity. Increased Ch content inhibits the insertion of A␤ into the membrane, resulting in an increase in A␤ concentration at the membrane surface and concomitant enhancement in the rate of A␤ fibrillogenesis (5). Therefore, the absence of Ch in the membrane will increase its fluidity, and so facilitate the insertion of the hydrophobic part of A␤ into the membrane (25). In Ch-free liposomes, the accumulation of A␤ into the membrane leads to its solubilization, resulting in the slow formation of thick fibril structures.
In the absence of both Ch and GM1, the characteristic CD pattern of ␤-structure with a negative at around 116 nm was not seen, and no fibril structure was observed by TEM. This suggests the coexistence of Ch and GM1 in the CCM membrane promote fibril formation. A strong decrease in turbidity at 40 M A␤ was also observed. All of these results suggest that the Ch-GM1-free membrane was solubilized by A␤ . Yanagisawa et al. (6) have shown that GM1 ganglioside-bound amyloid ␤-proteins are a possible form of preamyloid in AD. They also reported that oligomeric A␤ can promote the release of lipid from neurons to form A␤-lipid particles consisting of Ch, GM1, phospholipid, and A␤. Recent model membrane studies using liposomes consisting of GM1, Ch, and sphingomyelin showed that an increase in GM1 as well as Ch changes the binding capacity of A␤ (7,35). Our data indicate that GM1 and Ch strongly participate in the release of A␤ from the membrane and therefore are instrumental in amyloid precipitation.
It has been suggested that sphingolipids and cholesterol may exist as phase-separated "rafts" in sphingolipid and cholesterol-rich membranes such as the plasma membrane (36). The partial liquid-ordered rafts can be visualized as floating within the predominantly liquid crystalline "sea" of the lipid bilayer. Interestingly, it was proposed that the raft could be the site for the proteolytic processing of Alzheimer's amyloid precursor protein (APP) (37). Recently, we showed that elevating levels of sphingolipid and Ch cause decreased membrane fluidity and resulted in lipid-protein separations into liposomes containing ␣-helical transmembrane peptides (38). The proteolytic cleavage of APP to yield A␤ performed by both ␤-secretase and ␥-secretase present in the raft may in fact involve the release of A␤-lipid particles consisting of Ch, GM1, and phospholipid (8). However, in Ch-and GM1-free membranes, which are more fluid, A␤ is able to stay in the membrane, probably by insertion of its hydrophobic part into the lipid bilayer. The accumulation of A␤ in the membrane may result in its solubilization and eventual disruption into small vesicles.
The CCM consists of about 10% (w/w) plasmalogen (24), but the instability of this component under acidic conditions prevented its use in this study. We note, however, that it has been reported in the literature that levels of plasmalogen in AD CCM are reduced (39,40).
A␤-(1-42) has been recognized to be the more amyloidgenic component in plaques, since it has a greater propensity to form ␤-structure than A␤-(1-40), a requirement for amyloid fibral formation. Consequently, aggregates of A␤-(1-42) may act as an initiation factor for early plaque formation (10). In the present study, A␤-(1-40)/-(1-42) (10:1, molar ratio) formed ␤-structure more readily than A␤-  in all classes of liposome. Especially, in Ch-GM1-free membranes, where A␤-  forms no ␤-structure, the mixture of A␤-(1-40)/-  caused the gradual conversion of a predominantly ␣-helical structure into a mainly ␤-structure. Therefore, the presence of a small amount of A␤-  can induce the formation of ␤-structure in A␤-(1-40) and confirms the "seeding" hypothesis. A␤-(1-42) is able to promote the formation of ␤-structure that accompanies the formation of fibrils.
Recent studies have demonstrated that amyloid plaque formation may be initiated in the plasma membrane and that deposits were associated with the extracellular leaflet of the plasma membrane (41,42). Moreover, A␤-amyloid peptides are generated from various intracellular compartments, including the endoplasmic reticulum, the Golgi apparatus, lysosomes, and endosomes. The present model studies are carried out at the extracellular and lysosomal/endosomal pH values of 7.4 and 5.5. The results for the CCM liposomes indicate a kinetic pH-dependence of fibril formation, but it is not clear whether the release of A␤-(1-40) is also pH-dependent. The results seen for the Ch-GM1-free liposomes are the same in both pH environments; fibril formation does not occur. This definitely indicates that lipid bilayer composition plays an important role in the release of A␤ and might suggest that this release is much less dependent on the pH of the surrounding cytosol.
It has been shown that ganglioside and Ch participate in the mechanism of amyloid deposition in the presence of total brain lipid extract (5,11,34). However, until now there has been no report in the literature on the behavior of A␤ in membranes free of both Ch and GM1. We have shown that Ch and GM1 play an important role in the release of A␤ from liposome membranes designed to model cerebral cortex membranes, where fibril structure formation is known to be at its highest. In natural brain membranes, the A␤ generated from the processing of APP may be easily released from the membrane to play its correct biological role. However, the change of lipid composition in membranes by aging or other biological processes induces A␤ accumulation in membranes, this leads to formation of amyloid fibers or lipid-peptide particles, which are released into the cytosol, resulting in amyloid precipitation or cytotoxicity.