Cellular Membrane Composition Defines Aβ-Lipid Interactions

Alzheimer's disease pathology has demonstrated amyloid plaque formation associated with plasma membranes and the presence of intracellular amyloid-β (Aβ) accumulation in specific vesicular compartments. This suggests that lipid composition in different compartments may play a role in Aβ aggregation. To test this hypothesis, we have isolated cellular membranes from human brain to evaluate Aβ40/42-lipid interactions. Plasma, endosomal, lysosomal, and Golgi membranes were isolated using sucrose gradients. Electron microscopy demonstrated that Aβ fibrillogenesis is accelerated in the presence of plasma and endosomal and lysosomal membranes with plasma membranes inducing an enhanced surface organization. Alternatively, interaction of Aβ with Golgi membranes fails to progress to fibril formation, suggesting that Aβ-Golgi head group interaction stabilizes Aβ. Fluorescence spectroscopy using the environment-sensitive probes 1,6-diphenyl-1,3,5-hexatriene, laurdan,N-ε-dansyl-l-lysine, and merocyanine 540 demonstrated variations in the inherent lipid properties at the level of the fatty acyl chains, glycerol backbone, and head groups, respectively. Addition of Aβ40/42 to the plasma and endosomal and lysosomal membranes decreases the fluidity not only of the fatty acyl chains but also the head group space, consistent with Aβ insertion into the bilayer. In contrast, the Golgi bilayer fluidity is increased by Aβ40/42 binding which appears to result from lipid head group interactions and the production of interfacial packing defects.

Alzheimer's disease is an age-related disorder that is characterized by progressive cognitive decline and neurodegeneration (1,2). Pathological examinations have demonstrated that one of the key features is the presence of amyloid plaques associated with neuritic degeneration. Senile plaques are composed predominantly of a 40 -42-residue peptide, amyloid-␤ (A␤40/42). The development of Alzheimer's disease pathology has been proposed to be the result of A␤ 1 deposition in associ-ation with membrane structures. Recent studies have demonstrated that plaque formation may be initiated in a plasma membrane form (3,4) and that A␤ deposition in aged dogs is associated with the extracellular leaflet of the plasma membrane (5). Furthermore, the intracellular accumulation of A␤ in lysosomal or late endosomal vesicles in vitro suggest that these compartments may be involved in neurotoxicity (6 -9).
A␤ is generated from the proteolytic cleavage of the amyloid precursor protein in the endoplasmic reticulum to generate A␤42 and the trans-Golgi network to generate A␤40 (10 -14). It has also been suggested that A␤40/42 may also be generated at the plasma membrane surface. The presence of A␤ in distinct compartments and the proposal that lipid association is important for both neurotoxicity and fibrillogenesis suggest that the lipid composition and characteristics of these compartments may play vital roles in the disease process. Previous studies (15,16) have demonstrated accumulation of A␤42 in lysosomal compartments that results in membrane damage as shown by release of lysosomal hydrolases and the lysosomal specific dye acridine orange. Furthermore, A␤-synaptic plasma membrane interactions demonstrate that A␤ has a fluidizing effect on membrane structure as a result of A␤ insertion into the fatty acyl chain region of the bilayer (17). The role of proteins associated with synaptic plasma membranes could not be distinguished from A␤-lipid interactions alone in this study.
Therefore, we undertook the examination of A␤40 and A␤42 in the presence of bilayers formed from lipids isolated from post-mortem human cortical gray matter. We chose to evaluate the membranes involved in both the production of A␤, Golgi and endosomal, and A␤ pathology, plasma and lysosomal membranes. In order to distinguish between A␤ lipid and A␤ protein interactions in these compartments, we extracted the lipid component and used this as our model membranes. The effects of A␤ were examined as a consequence of sequence, structure, and concentration, all of which are factors affecting A␤ assembly and neurotoxicity. In order to address potential mechanisms to help explain the pathological findings, we examined the ability of these membranes to facilitate A␤40/42 assembly into amyloid fibers by electron microscopy. Changes in the membrane physical characteristics as a result of A␤ interactions were followed by fluorescence spectroscopy using environment-specific probes.

MATERIALS AND METHODS
Peptides-A␤40/42, A␤-  were synthesized by solid phase Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry by the Hospital for Sick Children's Biotechnology Center (Toronto, Ontario, Canada). They were purified by reverse phase high pressure liquid chromatography on a C18 Bondapak column. A␤ peptides were initially dissolved in 0.5 ml of 100% trifluoroacetic acid (Aldrich), to ensure that the peptide re-mained monomeric and free of fibril seeds, diluted in distilled H 2 O, and immediately lyophilized (18). A␤ peptides were then dissolved at 1 mg/ml in 40% trifluoroethanol (Aldrich) in distilled H 2 O and stored at Ϫ20°C until use. Bee venom mellitin was used as a control peptide (Sigma).
Cellular Membrane Isolation-All cellular membranes were isolated from post-mortem human gray matter of five male control subjects with post-mortem intervals of less than 15 h. The male subjects ranged in age from 76 to 80 years without documented signs of clinical dementia. The cause of death in all cases was heart failure. Plasma membranes were isolated using the method of Hubbard et al. (19), endosomes using the method of Gorvel et al. (20), and Golgi membranes isolated using the method of Duden et al. (21), all of which rely upon the separation of the specific fraction by differential migration in sucrose density gradients. Lysosomal membranes were isolated using the procedure of Storrie and Madden (22) using flotation on a metrizamide density gradient. Lipids were extracted from each membrane fraction using chloroform:methanol (2:1) extraction and subsequent concentration under a stream of N 2 . The samples were stored at Ϫ20°C until use. Phospholipid concentration in all samples was determined using the Bartlett assay (23), and cholesterol concentration was determined using the Amplex red assay (Molecular Probes, Eugene, OR).
Electron Microscopy-A␤40/42 peptides were incubated in the presence and absence of total brain lipid extract bilayers at a final peptide concentration of 100 g/ml. The A␤ to lipid ratio was maintained at 1:20 (by weight). For negative stain electron microscopy, carbon-coated pioloform grids were floated on aqueous solutions of peptides. After the grids were blotted and air-dried, the samples were stained with 1% (w/v) phosphotungstic acid and examined on a Hitachi 7000 electron microscope operated at 75 kV (29,30).
Steady State Fluorescence Anisotropy-Anisotropy experiments were performed on a PTI fluorimeter equipped with manual polarizers as described previously (24). Excitation and emission wavelengths were set at 360 and 425 nm with a slit width of 1 and 4 nm, respectively. Our system was initially calibrated using 1,6-diphenyl-1,3,5-hexatriene (DPH; Molecular Probes, Eugene, OR) in mineral oil, which should give an anisotropy equal to 1. The g factor was calculated using horizontally polarized excitation and subsequent comparison of the horizontal and vertical emissions, which for our machine is 0.883. Lipid vesicles were diluted to 250 g/ml in phosphate-buffered saline, incubated for 20 -30 min in the presence and absence of A␤, and then subsequently incubated for a further 30 min with DPH at a 1:500 probe:lipid ratio. Fluorescence intensity was measured with the excitation polarizer in the vertical position and the analyzing emission polarizer in the vertical (I VV ) and horizontal (I VH ) positions; and anisotropy, r, was calculated using Equation 1, Lipid vesicles in the absence of DPH were measured in order to evaluate the effect of light scattering on our measurements. Poly-L-lysine and bovine serum albumin were used as negative controls for the anisotropy studies.
Laurdan Generalized Polarization-Steady state excitation and emission spectra were collected on the PTI fluorimeter. Laurdan (Molecular Probes) was added to preformed lipid vesicles in the presence and absence of A␤ at a 500:1 lipid:probe ratio. The laurdan generalized polarization (GP) parameter as developed by Parasassi et al. (25) is calculated as follows. The emission GP parameter is given by Equation 2.
where I 400 nm and I 340 nm are the fluorescence intensities measured at all emission wavelengths within 420 and 520 nm. By using fixed excitation wavelength of 400 nm and 340 nm, respectively. The excitation GP is given by Equation 3, where I 440 nm and I 490 nm are the fluorescence intensities at each excitation wavelength from 320 to 420 nm, measured at fixed emission wavelengths of 440 and 490 nm, respectively.
(DL, Molecular Probes) was incorporated into lipid vesicles in the presence and absence of A␤. The fluorescence spectra of DL were evaluated after 30 min of incubation at room temperature with an excitation wavelength of 335 nm and emission scan monitored between 380 and 580 nm inclusive. The DL to lipid ratio was maintained at 1:500 (26).
Merocyanine 540 Absorption Spectroscopy-Merocyanine 540 (MC540, Molecular Probes) absorption spectra were obtained at room temperature on a Beckman spectra DU530. The dye was added to preformed vesicles at a probe:lipid ratio of 1:500 (27). Final MC540 molar concentration in the cuvette was 21.3 ϫ 10 Ϫ6 M. Absorption spectra were obtained between 400 and 600 nm with 1-nm steps. The lipid-alone base line in the absence of MC540 was subtracted from all spectra, and the corresponding spectra are shown in Equation 4, and were then corrected by referring the absorbances at 600 nm to 0. After this correction, the absorbance values at 569 nm were used to calculate the dimerization constant (K d(app) ) as by Bernik and Disalvo (28), see Equation 5.
where A is the absorbance at 569 nm, ⑀ is the constant for MC540 dimer or monomer at the given wavelength, ⑀ m ϭ 1.511 ϫ 10 5 and ⑀ D ϭ 5400, and C is the final MC540 concentration.

RESULTS
A␤ Morphological Characteristics-Lipid bilayers have been shown to affect the assembly of A␤ peptides into amyloid fibers (29 -31). In order to determine if A␤ interactions with different cellular membranes affects fibrillogenesis, we examined A␤ structural characteristics in the presence of vesicles formed from Golgi, plasma, lysosomal and endosomal lipids by negative stain electron microscopy. In the absence of lipid, A␤ assembles into long fibers of varying length, 350 -430 Å, with a characteristic helical twist of 100 Å (Fig. 1A). These fibers demonstrated varying extent of lateral aggregation of fibers into larger bundles, from 50 Å representing single fibers to 200-Å diameter bundles. In the presence of plasma lipid vesicles, A␤ assembled into fiber bundles along the surface of the bilayer (Fig. 1B). A␤ fibers were not found on the surface of the When incubated in the presence of plasma membrane (B), a similar structure of the fibrils to A␤42 alone could be detected but with increased organization along the vesicle surface. Only minor lateral aggregation was apparent in the fibrils formed in the presence of lysosomal membranes (C). In the presence of Golgi membranes only a few protofibrils of A␤42 were detected (D). Scale bar is 50 nm.
bilayers nor in areas devoid of lipid vesicles, suggesting that A␤ assembly was driven as a result of interaction with the lipid surface. These results are similar to those that we have reported previously for A␤ interaction with phosphatidylinositol/ brain phosphatidylcholine and total brain lipid vesicles (32,33). Both endosomal (data not shown) and lysosomal ( Fig. 1C) vesicles demonstrated A␤ fibers associated with both the edges and surface of the vesicles. Although no fibers were detected in the absence of lipids, the level of A␤ lateral aggregation and organization was less than that detected for the plasma membrane lipids. A␤ incubated in the presence of Golgi lipid vesicles was almost devoid of A␤ fibers (Fig. 6D). The odd fiber could be found across the grid but was not intimately associated with lipid vesicles, and many areas of lipid vesicles could be found devoid of fibers. The odd fiber detected in the presence of Golgi vesicles had morphological characteristics of A␤ protofibrils. The Golgi lipid vesicles are reminiscent of our previous results in which we were unable to identify A␤ fibrils in the presence of ganglioside/phosphatidylcholine membranes when A␤ was added as a randomly structured peptide (29).
Fatty Acyl Chain Mobility-In order to characterize differences in the cellular bilayer properties and determine which is most influential in determining A␤ fibrillogenesis, we examined the influence of A␤40/42 on the physical properties of these bilayers. Many fluorescent dyes are available that penetrate to varying levels into the lipid bilayer and exhibit fluorescent properties indicative of their local environment. We can utilize these dyes to address the effects of A␤-lipid interactions within various cellular membranes. Previous studies using synthetic lipid bilayers and synaptic plasma membranes have demonstrated a disordering of the fatty acyl chains after interaction with A␤40 and A␤42 (17, 34 -36). In order to determine the effect of A␤40/42 on the mobility of the fatty acyl chains within bilayers formed from Golgi, endosomal, lysosomal and plasma membrane lipids, we have examined the steady state fluorescence anisotropy using the dye, DPH (24). The relative motion of the DPH dye molecule within the lipid bilayer is determined by polarized fluorescence and expressed as r, the anisotropy constant. This constant is inversely proportional to the degree of membrane fluidity.
The relative fluidity of the lipid membranes was found to vary considerable with Golgi lipid bilayers having the most rigid structure, whereas plasma membrane, endosomal, and lysosomal lipid bilayers were more fluid (Table I). Previous studies have shown that synthetic lipid bilayer fluidity is regulated by the amount of cholesterol, which exhibits a bimodal effect on fluidity with increasing levels of cholesterol (34,(37)(38)(39). The cholesterol to phospholipid ratios of the bilayers were in the range of that reported previously in the literature (Table  II). The differences in fluidity detected in this study are not solely related to the cholesterol content of these bilayers as the plasma and lysosomal membranes have a significant choles-terol level yet still exhibited a fluid membrane structure. Therefore, the properties of the different bilayers represent differences in the total lipid composition rather than cholesterol per se.
Previous reports (17, 34 -36, 39) have shown that the addition of A␤40 or A␤25-35 to synthetic membrane preparations or membranes isolated from red blood cells results in a reduction in membrane fluidity. Many A␤ properties have been linked to the conformation and aggregation state of the peptide. In order to investigate the interactions of A␤40 and A␤42 with lipid bilayers, we chose to examine initially soluble, random structured peptide with bilayers. To ensure that A␤ peptides meet these criteria and are free of fibril nucleation seeds, A␤40/42 peptides were treated with 100% trifluoroacetic acid followed by lyophilization (18). The lyophilized peptide was immediately solubilized in 40% trifluoroethanol in order to make a 1 mg/ml stock solution. As reported previously (40), A␤40/42 is partly ␣-helical in 40% trifluoroethanol and upon dilution into phosphate-buffered saline, pH 7.4, A␤40/42 initially adopted a random structure. Our results demonstrate that addition of randomly structured A␤40 and A␤42 decreased the membrane fluidity of the plasma membrane, endosomal and lysosomal membranes in a concentration-dependent manner as illustrated by an increase in the anisotropy constant (Table I and Fig. 2). We could not detect a significant difference in the effect of A␤40 and A␤42 on these membranes. In contrast, both A␤40 and A␤42 induced a significant increase in fluidity of the Golgi lipid bilayers as demonstrated by the decrease in the anisotropy constant (Table I). The disordering effect of A␤ on Golgi lipid membranes is enhanced by increasing A␤ concentration (Fig. 2).
To determine the specificity of A␤40/42-lipid interactions, we examined A␤-(1-28), which lacks the N-terminal hydrophobic region, and bee venom mellitin, a pore-forming peptide. Neither A␤-(1-28) nor mellitin altered the fluidity of the Golgi membranes, whereas both decreased the fluidity of plasma membranes (Table I). These results suggest that the preferential increase in Golgi membrane fluidity as a result of A␤40/42 interaction is peptide-and sequence-specific. It is not surpris-  ing that mellitin decreases the fluidity of plasma membranes as it inserts into membranes to create pores. The structural dependence of A␤ effects on membrane fluidity of plasma membrane and Golgi lipid bilayers was examined by comparing the random structured peptide with A␤40 and A␤42 which exhibit ␤-structure. In contrast to the random structured peptides, seeded A␤40 decreased the membrane fluidity of both plasma and Golgi membranes (Table I). Similar results were detected for A␤42. This result suggested to us that A␤ interactions with lipid bilayers is not only dependent on the composition of the lipid bilayer but also on the structural characteristics of the peptide.
Dynamics of Lipid Head Groups and Interface-Besides the packing of the lipid acyl chains, the dynamics of the polar head groups and the polarity of the lipid interface are relevant to the interaction of molecules, i.e. A␤, with the membrane surface. In order to obtain an insight into these properties, laurdan and N-⑀-dansyl-L-lysine probes were used. Laurdan naphthalene ring is located at the glycerol backbone and is anchored in the bilayer by the lauroyl moiety, thereby imparting fluorescence characteristics that are dependent on the polarity of its environment (25,41). The advantages of laurdan are that it is completely non-fluorescent in aqueous environments, is independent of pH between 4 and 10, and independent of lipid polar head group; therefore fluorescence readings reflect only the polarity of the probe associated with the bilayer. The spectral properties of laurdan have been described by the general polarization equation for both excitation and emission spectra, which render information about the lipid phase, polarity, and co-existence of multiple lipid phases within a single bilayer (25,41).
Laurdan excitation spectra in the presence of plasma and Golgi lipid bilayers demonstrate the characteristic red excitation at 340 nm and blue excitation at 380 nm, whereas the emission spectra indicates a single maximum at 430 nm indicative of blue emission (Fig. 3). The red excitation band intensity increases in polar solvents, and in hydrogen-bonding solvents, the red excitation corresponds to the blue emission population and is especially intense in gel phase lipid bilayers where little relaxation occurs. The addition of A␤ to laurdan containing membranes does not change the shape of either the excitation or emission spectra but affects the intensity of laurdan fluorescence in both plasma and Golgi membranes (Fig. 3). The ratio of the blue to red components in the excitation reflects the polarity of the probe. Addition of A␤40 to plasma membrane bilayers results in an increase in the blue/red excitation ratio from 1.03 to 2.2, indicating that the environment sensed by laurdan becomes more hydrophobic after interaction of A␤ with membranes (Fig. 2). These results suggest that A␤ produces a displacement of water molecules from the hydration shell of the membrane, as a result of the promotion of lateral phase separation and a higher degree of plasma membrane organization. Increasing the concentration of A␤ does not further alter the blue/red excitation ratio suggesting that the bilayer has a finite ability to accommodate A␤. The generalized polarization emission (GP em ) for plasma membranes was calculated to be 0.48, which increases to 0.57 in the presence of initially random structured A␤40 and A␤42. The interaction of seeded A␤40/42 demonstrates the same shift of the GP em to 0.56 and 0.54, respectively, indicating that in the presence of A␤ the membrane becomes more structured at the head group-fatty acyl chain interface. In contrast, both Golgi (Fig. 3B) and endosomal (data not shown) bilayer blue/red excitation intensities do not change after addition of A␤ suggesting that A␤ binding does not alter the phase of these lipids. Increasing concentrations of A␤ still did not induce a change in the ratio of excitation intensities confirming that these bilayers do not undergo a concentrationdependent phase transition. The endosomal lipid bilayer exhibits similar fluidity as the plasma membrane as illustrated by DPH studies, yet has different laurdan fluorescence characteristics suggesting that bilayer-specific lipid composition may alter the resultant A␤-lipid interactions. In contrast to the excitation properties of laurdan, the GP em values of Golgi and endosomal membranes, 0.49 and 0.54, are only affected by the addition of high concentrations of A␤ demonstrating a decrease to 0.42 and 0.44 for A␤40. It is interesting that A␤42 has a lesser effect on the GP em than A␤40 with only modest alteration of the GP em values to 0.47 and 0.51, respectively. These results suggest that interaction of A␤42 with both Golgi and endosomal membranes does not alter the micropolarity or hydration of the interfacial region of the lipid bilayer.
To examine the specificity of the changes in laurdan fluorescence properties due to A␤40/42 interactions with Golgi and plasma membranes, we examined the effects of A␤-(1-28) and mellitin. Addition of A␤-(1-28) and mellitin to both bilayers does not alter the blue/red excitation ratio. Furthermore, A␤-(1-28) does not shift the GP em of either plasma or Golgi membranes. These combined results suggest that A␤-(1-28) does not alter the micropolarity of the interface and/or insert into the bilayer. In contrast, mellitin shifts the GP em for Golgi and plasma membranes from 0.49 and 0.48 to 0.53 and 0.52, respectively. These results suggest that in contrast to A␤ peptides, mellitin binding and pore formation increases the lipid order after insertion into the bilayer but does not change the lipid phase.
In order to examine more closely the phase behavior of the bilayers, the wavelength dependence of both the excitation and emission spectra was examined (Fig. 4). Lipid bilayers in a pure gel phase show an independence of generalized polarization values as a function of wavelength, whereas liquid crystalline bilayers exhibit a dependence on the excitation wavelength (42). Plasma membrane bilayers exhibit a decrease in the GP ex and increase in GP em toward shorter wavelengths, an indication of the co-existence of lipid phases (Fig. 4A). The addition of A␤ results in an increase in the slope of the GP em suggesting that the lipid phase in the bilayer is further altered as a result of A␤ binding. Alternatively, the GP ex and GP em of Golgi lipid bilayers is independent of wavelength in the presence and absence of A␤, confirming that A␤ binding does not alter the lipid phase (Fig. 4B). These results suggest that binding of A␤ to Golgi and endosomal membranes can be easily accommodated within the lipid structure, whereas plasma membrane bilayers undergo a reorganization.
The polarity of the lipid interface can be examined using the fluorescence of N-⑀-dansyl-L-lysine (26,43); furthermore, it has been suggested that DL inserts into cholesterol-free phospholipid domains (44,45). Due to its molecular structure and location at the interface, DL fluorescence is most sensitive to the packing constraints and hydration. DL exhibits a strong fluorescence maximum at 430 nm, which increases in intensity as a result of A␤-plasma membrane interactions (data not shown). These results suggest that A␤ increases the polarity of the interface which is independent of concentration and structure. In contrast, both endosomal and Golgi lipid bilayers exhibit DL maxima at 430 and 540 and 520 nm, respectively. The addition of A␤40 and A␤42 to endosomal bilayers results in a blue shift in the DL maxima; an indication of increased lipid packing and was independent of peptide structure (data not shown). DL associated with Golgi membranes demonstrated a red shift in the fluorescence maxima after addition of A␤. This result suggests that A␤ creates more space between the lipid head groups or causes an increase in the packing defects of Golgi lipid bilayers.
Lipid Head Group Packing and Surface Properties-In order to examine the lipid head group spacing and surface properties of these bilayers, merocyanine 540 absorbance spectral properties were examined. The spectral characteristics of MC540 result from binding of monomeric MC540 and subsequent dimerization, and both steps are dependent on the packing properties of the lipid head groups (27,28). MC540 spectra in the presence of plasma membrane is characteristic of mostly gel phase lipid head groups, with the characteristic maxima at 500 and 530 nm (Fig. 5A). A small shoulder is present at 570 nm which is characteristic of a small population of monomeric MC540 insertion into the lipid bilayer. These results suggest an ordered head group packing in these bilayers as only a small amount of MC540 is inserted into the head group space. Addition of A␤40 to the plasma lipid bilayers decreased the intensity of the MC540 maxima and percent of monomeric MC540 insertion (Fig. 5B). No difference could be detected between random and ␤-structured A␤40 suggesting similar effects on lipid head group rearrangement. A␤42 did not change the absorbance spectra of MC540, suggesting that A␤42 interaction does not affect the head group packing of the plasma membrane. On the other hand, similar MC540 spectra results were obtained for endosomal lipid bilayers in the presence of A␤ (data not shown). The MC540 spectra in the presence of plasma and endosomal membranes are consistent with varying levels of A␤ insertion into the bilayers.
These results are contrasted by the MC540 spectra in the presence of Golgi lipid bilayers, which demonstrate maxima at 530 and 570 nm (Fig. 5A). These spectra are indicative of a more fluid, liquid-crystalline head group packing and an increased surface potential that allow for increased MC540 monomeric insertion into the head group space. Addition of A␤40 and A␤42 results in an increase in the intensity of both maxima, indicating a more fluid environment and increased head group space or packing defects (Fig. 5C). In contrast, seeded A␤40 and A␤42 decrease the intensity of the 570 nm maxima suggesting that ␤-structured peptide increases the packing of the head groups of Golgi membranes. The MC540 absorption spectra are consistent with A␤40/42-Golgi interactions occurring predominantly at the head group space.
To investigate the sequence specificity of A␤40/42 interaction with Golgi and plasma membrane bilayers, we examined the interaction of A␤-(1-28) under similar conditions. In contrast to A␤40/42, A␤-(1-28) did not affect the shape or intensity of the MC540 absorption spectra of Golgi membranes. These results suggest that A␤-(1-28) does not affect head group packing and confirms the DPH and laurdan fluorescent results, which suggest that A␤-(1-28) does not insert into the lipid bilayer (data not shown). Similar to A␤40/42, mellitin increases the intensity but not the shape of the MC540 absorption spectra. Furthermore, MC540 spectra of plasma membranes in the presence of mellitin are indistinguishable from that of Golgi membranes (Fig. 5D). These results are consistent with mellitin insertion into the bilayer and creating increased head group space or packing defects.
The MC540 monomer-dimer equilibrium is relevant to the packing properties of the bilayers and can be used as an indication of lipid head group spacing (27,28). We have calculated the apparent dimerization constant for plasma, Golgi, and endosomal lipid bilayers in the presence and absence of A␤40/42 in both random and ␤-structure (Table III). The most apparent observation is that the dimerization constant for the various bilayers differs on the order of 2 magnitudes from each other in the order Golgi Ͻ plasma Ͻ endosomal bilayers. These results suggest that the head group packing of the Golgi membranes is less constrained and can accommodate the MC540 dimers.
Furthermore, addition of A␤40/42 did not significantly alter the K d (app) suggesting that the membranes can easily accommodate A␤. Our anisotropy studies suggest that both the plasma and endosomal lipid bilayers are both fluid bilayers, whereas the dimerization constant suggests that the endosomal head group packing is more rigid than the plasma membrane bilayers. Addition of A␤40/42 as a randomly structured peptide did not alter the K d(app) , suggesting that A␤ binding does not alter head group packing. The dimerization of MC540 in endosomal lipid bilayers was decreased by A␤ binding as indicated by a 3-fold increase in the dimerization constant (Table III) after addition of both random and ␤-structured peptides. These results suggest that A␤-endosomal interactions further organize the head group packing and are consistent with our anisotropy studies, which demonstrate a decrease in the fatty acyl chain fluidity as a result of A␤ binding to endosomal bilayers. DISCUSSION A␤-lipid interactions have implications not only for A␤ production but also for the induction of neurotoxicity and ageassociated pathology. The presence of A␤ aggregates, initiation of plaque formation, and the dependence of toxicity on the association with specific lipid compartments suggested that vesicular lipid composition might be a factor in these processes. Our fluorescence studies on plasma membrane bilayers suggest that A␤ inserts into the fatty acyl region of the bilayer. This result is supported by the anisotropy studies that demonstrate a dramatic increase in membrane organization as a result of A␤ interaction, the lipid phase shift associated with A␤ as demonstrated by the laurdan GP and the lack of head group reorganization as detected by MC540 absorption spectra. Our results are consistent with previous reports (34, 36) that demonstrated FIG. 5. The interaction of A␤40 and A␤42 with the lipid head groups of the various cellular membranes was examined using MC540 absorbance spectroscopy. MC540 spectra demonstrate the rigid packing of the plasma membrane head groups (A, dashed line), whereas the Golgi membrane head groups are more fluid (A, solid line). Addition of A␤40 (dashed line) and A␤42 (dotted line) to Golgi membranes (C) resulted in an increase in the intensity of the MC540 spectra indicative of A␤-head group interactions. In contrast, A␤42 had little effect on the plasma membrane (B) as indicated by a lack of shift in the spectra, whereas A␤40 decreased the intensity of the spectra indicating increased packing of the head groups. In contrast, mellitin (D, dashed line) in the presence of plasma membrane (D, solid line) shifted the shape of the MC540 spectra to one that is indistinguishable from Golgi membranes, suggesting creation of lipid head group packing defects or altered spacing. that addition of A␤ to membranes isolated from cerebellum, cortex, hippocampus, and striatum or synthetic lipid vesicles results in a decrease in membrane fluidity. In contrast, Mason et al. (17) reported that A␤, both random and aggregated, increased synaptic plasma membrane fluidity by insertion of random A␤ into the fatty acyl chain and the presence of aggregated A␤ at the lipid head groups. Our results are in partial agreement with these results, as we also propose that A␤ exerts its bilayer effects by inserting into the fatty acyl chains, but differ in that our results demonstrate a rigidizing effect. The discrepancies between these two studies may be accounted for by the presence of endogenous synaptic membrane proteins that may compete with lipids for A␤ binding. Our results further demonstrate the enhancement and organization of A␤ fibrillogenesis in the presence of plasma membrane vesicles. These structural results are consistent with our previous studies that demonstrated the insertion and fibrillogenesis of A␤40 on planar bilayers composed of total brain lipid extracts (33) and the presence of fibers in synthetic phospholipid bilayers as detected by electron microscopy (32). The ability of A␤ to insert into the plasma membrane has many implications for both cell survival and cell surface-driven fibrillogenesis. Endosomal and lysosomal lipid bilayers have similar properties to plasma membranes except that upon interaction with A␤ the lipid head groups undergo re-organization. Endosomal head group organization was initially demonstrated to be rigid which may limit the level of A␤ insertion into the lipid bilayer. The endosomal compartment is the site of cholesterol uptake and recycling within the cell, and the limited A␤ insertion into the bilayer may result from increased cholesterol to phospholipid ratio. Previous studies have demonstrated that cholesterol modulates A␤-lipid interactions by preferential binding, decreasing the fluidity of the bilayer, and ultimately decreasing fiber and aggregate formation (34,35,39). Furthermore, the lysosomal and endosomal compartments have been suggested to be sites of intracellular A␤ accumulation and nucleation. The increased packing of the endosomal head groups would suggest that accumulation of A␤ would be near the surface of the bilayer, a site that would be easily accessible for propagation of A␤ nucleation and aggregation.
We have demonstrated that the interaction of A␤ with Golgi membranes is predominantly at the level of the head groups but also translates into a decreased micropolarity at the head group-fatty acyl chain interface and decreased order of the fatty acyl chains. Our studies have demonstrated that Golgi membranes can easily accommodate A␤ and ultimately inhibit fiber formation. These results may be due to the high glycolipid concentration in these bilayers since our previous studies (29) on isolated brain ganglioside-A␤ interactions demonstrate similar inhibition of fibrillogenesis. The mechanism of action was proposed to be surface binding in an ␣-helical conformation that prevented conversion to ␤-structure and subsequent fibrillogenesis. Similarly, the laurdan fluorescence characteristics of Golgi membranes alone are characteristic of those reported previously (46,47) for glycosphingolipid containing aggregates and vesicles. A␤40 is generated in the Golgi apparatus, and it would not be beneficial to cell survival for the Golgi to possess properties that promote A␤ self-aggregation. Furthermore, it has been suggested that plasma membrane-generated A␤ occurs in membrane rafts that are rich in glycolipids. These data may represent a protective mechanism against A␤ toxicity.
Our results demonstrate that the cellular vesicular compartments exhibit lipid characteristics that either promote or inhibit fibril formation by direct interaction with lipid bilayer. Although we have examined A␤-lipid interactions of various compartments, we have not taken into account the effect of endogenous cellular proteins. These integral membrane proteins will also have an effect on A␤-membrane interactions whether as competitors for A␤ binding, such as proteoglycans, or as modulators of bilayer properties. Our results demonstrate differences detected in the A␤-lipid interactions between the various vesicular compartments, which may play a role in not only normal cellular processing and turnover of A␤ but in the progression of disease processes in Alzheimer's disease.