High Membrane Curvature Enhances Binding, Conformational

Oct 16, 2015 - Aggregation of the amyloid-β (Aβ) protein and the formation of toxic aggregates are the possible pathogenic pathways in Alzheimer's d...
0 downloads 6 Views 2MB Size
Download PDF
Article pubs.acs.org/Langmuir

High Membrane Curvature Enhances Binding, Conformational Changes, and Fibrillation of Amyloid‑β on Lipid Bilayer Surfaces Yuuki Sugiura, Keisuke Ikeda,* and Minoru Nakano Graduate School of Medicine and Pharmaceutical Sciences, University of Toyama, 2630 Sugitani, Toyama 930-0194, Japan S Supporting Information *

ABSTRACT: Aggregation of the amyloid-β (Aβ) protein and the formation of toxic aggregates are the possible pathogenic pathways in Alzheimer’s disease. Accumulating evidence suggests that lipid membranes play key roles in protein aggregation, although the intermolecular forces that drive the interactions between Aβ-(1−40) and the membranes vary in different membrane systems. Here, we observed that a high positive curvature of lipid vesicles with diameters of ∼30 nm enhanced the association of Aβ with anionic phosphatidylglycerol membranes in the liquid-crystalline phase and with zwitterionic phosphatidylcholine membranes in the gel phase. The binding modes of Aβ to these membranes differ in terms of the location of the protein on the membrane and of the protein secondary structure. The fibrillation of Aβ was accelerated in the presence of the vesicles and at high protein-to-lipid ratios. Under these conditions, the protein accumulated on the surfaces, as demonstrated by a high (107 M−1) binding constant. Our findings suggest that packing defects on membranes with high curvatures, such as the intraluminal vesicles in multivesicular bodies and the exosomes, might result in the accumulation of toxic protein aggregates.

■

fibrils.16−20 It was suggested that Aβ aggregates formed in the presence of membrane components exhibited higher cytotoxicity than the ones generated in the absence of lipids.21,22 Recent studies have also demonstrated that Aβ accumulated in specific intra- or extracellular membrane structures. Takahashi et al. reported that in healthy mice, rats, and humans, Aβ-(1−42) localized to multivesicular bodies (MVBs) of neuronal cells. MVBs are late endosomes that contain intraluminal vesicles with a diameter that ranges from 30 to 80 nm. They also showed that, in transgenic mice and in the brains of patients with Alzheimer’s disease, Aβ-(1−42) accumulated in MVBs with aging.23 Friedrich et al. have observed that upon addition of Aβ to THP-1 and J774A.1 cells, the protein accumulated within MVBs, forming fibrils that disrupted the membrane structures and induced cell death.24 Rajendran et al. indicated that Aβ was secreted from HeLa and N2a cells through the exosomes, which are intraluminal vesicles that form upon fusion of MVBs with the plasma membrane and are subsequently released in the extracellular space.25 Yuyama et al. induced the release of exosome-associated GM1 by disrupting the endocytic pathway in PC12 cells and observed the acceleration of Aβ aggregation.26 These studies raise the possibility that Aβ preferentially binds to small intra- or extracellular vesicles with high membrane curvature, although the presence of specific lipid components such as GM1 ganglioside might also influence the interaction. Indeed, Terakawa et al. have recently reported that small lipid vesicles (≤50 nm) composed of dioleoyl- or palmitoyloleoyl-PC

INTRODUCTION Alzheimer’s disease is a neurodegenerative disorder characterized by the presence of senile plaques in the brains of patients.1 These plaques mainly consist of aggregates of the amyloid-β (Aβ) protein, a 39−43 residue protein derived from the proteolytic cleavage of the amyloid precursor protein. It is widely accepted that the conversion of Aβ, a nontoxic soluble monomer, into toxic aggregates such as small oligomers and amyloid fibrils is a crucial step in the pathology of the disease.2,3 Monomeric Aβ did not show neurotoxicity nor inhibit longterm potentiation.4,5 Rather, it exhibited antioxidant and neuroprotective effects.6,7 Aβ aggregates formed both in vitro and in vivo, however, have been reported to exert toxicity on neuronal and cultured cells and to induce neuronal loss and cognitive impairment in model animals.8−10 In the past few decades, a number of studies have identified a variety of factors, such as metal ions, sugars, and proteins, that interact with Aβ under pathological conditions, triggering its aggregation.11 However, accumulating evidence suggests that the interactions between lipid membranes and Aβ play a key role in the aggregation process.12,13 For example, the binding of Aβ to phosphatidylglycerol (PG) and 1,2-dioleoyl-3-trimethylammonium-propane membranes is driven by electrostatic interactions and is dependent on the pH and ionic strength.14 Yoda et al. reported the interaction between Aβ and electrically neutral phosphatidylcholine (PC) liposomes at low temperatures, when the lipid bilayers were in the gel phase.15 Several studies indicated that when Aβ is bound to GM1 ganglioside, a glycosphingolipid abundant in neuronal cell membranes, it undergoes a conformational change from random coil to αhelix- or β-sheet-rich structures, eventually forming amyloid © XXXX American Chemical Society

Received: June 5, 2015

A

DOI: 10.1021/acs.langmuir.5b03332 Langmuir XXXX, XXX, XXX−XXX

Article

Langmuir Table 1. Protein Sequences protein

amino acid sequence

Aβ-(1−40) Aβ[Y10W] Aβ[Y10W/H13K]

NH2-DAEFRHDSGYEVHHQKLVFFAEDVGSNKGAIIGLMVGGVV-COOH NH2-DAEFRHDSGWEVHHQKLVFFAEDVGSNKGAIIGLMVGGVV-COOH NH2-DAEFRHDSGWEVKHQKLVFFAEDVGSNKGAIIGLMVGGVV-COOH

accelerated Aβ fibrillation.27 However, the binding modes and the mechanisms of Aβ aggregation on lipid vesicles of high curvature are not fully understood. Here, we used two types of lipid membranes reported to interact with Aβ in order to investigate the influence of the membrane curvature on the binding, on the conformational changes, and on the aggregation kinetics of Aβ-(1−40), the most common form of Aβ. We showed that membranes with high curvature exhibited a higher affinity for Aβ and that they induced structural transitions of the peptide. The binding modes and the final structures adopted by Aβ were significantly different between the two membrane systems because of their specific physicochemical properties. We also observed that the high protein density on the surface of highly curved vesicles accelerated Aβ aggregation.

■

using an enzyme assay kit for choline (Wako, Osaka, Japan), and the concentration of PG was determined by phosphorus analysis.28

Table 2. Diameters of the Lipid Vesiclesa SUVs LUVs

DMPC

POPG

POPC

DSPC

31 ± 1b 129 ± 7b

30 ± 5b 107 ± 9b

31 ± 3b

57 ± 1b

a

Nanometers. bAverages and standard deviations of the mean diameters for duplicate or triplicate experiments. Tryptophan Fluorescence Titration. Fluorescence measurements were carried out on an F-4500 spectrofluorometer (Hitachi, Tokyo, Japan) equipped with a cuvette holder thermostated at 4 °C for DMPC or at 25 °C for POPG and POPC. Fluorescence emission spectra were recorded using an excitation wavelength of 280 nm. Aβ[Y10W] or Aβ[Y10W/H13K] solutions were diluted in 2 mL of Tris-HCl buffer to a concentration of 2.5 μM in a quartz cuvette and were left standing on the cuvette holder for 10 min before recording the spectra. The lipid/protein (L/P) ratio was increased upon titration with aliquots of the vesicle suspensions. The mixture was gently stirred for 3 min to reach the equilibrium conditions for the binding. We measured the spectrum of the vesicle suspensions as a blank and that of a mixture of L-tryptophan (2.5 μM, 2 mL) with the vesicles as a control for calibrating the scattered light. Each titration experiment was corrected for the blank and normalized using the control spectra. Circular Dichroism (CD) Spectra. CD spectra were recorded at 4 or 25 °C on a J-805 apparatus (Jasco, Tokyo, Japan) using a 5.0 mm path length quartz cuvette. The concentration of Aβ in Tris-HCl buffer was 2.5 μM. Eight scans were averaged for each sample. The averaged blank spectra (vesicle suspension or buffer) were subtracted. To avoid the absorbance of the chloride ions and to obtain better-quality spectra in the PC systems, the NaCl salt in the buffer was replaced with NaF.17 Thioflavin-T (ThT) Assay. Aβ (15 μM) was incubated with the vesicles in buffer at 37 °C without agitation. Fibrillation was monitored by ThT fluorescence.29,30 Aliquots of the incubated solution were sampled at various incubation times and diluted with 50 mM glycine buffer (pH 8.5) containing 5 μM ThT at a final Aβ concentration of 0.5 μM in a quartz cuvette. The fluorescence at 490 nm was measured using an excitation wavelength of 446 nm on an F-4500 spectrometer equipped with a cuvette holder thermostatically controlled at 25 °C. Transmission Electron Microscopy (TEM). The samples in which fibrillization was confirmed by the ThT assay were spread on carbon-coated grids, negatively stained with 2% phosphotungstic acid, and examined under a JEM-1400TC electron microscope (JEOL, Tokyo, Japan).

MATERIALS AND METHODS

Peptides. Aβ-(1−40), Aβ[Y10W], containing a Y to W mutation at position 10 on the hydrophilic domain of Aβ-(1−40), and Aβ[Y10W/ H13K], containing a H to K mutation at position 13 of Aβ[Y10W], were synthesized by the standard fluorenyl-methoxy-carbonyl (Fmoc)-based solid-phase method (Table 1). The crude peptides were purified by reverse-phase high-performance liquid chromatography (HPLC), and the solvent was removed by freeze-drying. Matrixassisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOFMS) was used to confirm the identity of the species. The purified peptides were dissolved in 0.02% NH3(aq) on ice. We removed any aggregates that might act as a seed for further aggregation by ultracentrifugation in 500 μL polyallomer tubes at 527 000g for 3 h at 4 °C.17 The supernatant was dispensed into aliquots, frozen with liquid nitrogen, and stored in a freezer at −80 °C. The peptide concentration in the supernatant was determined in triplicate by micro BCA protein assay. Before the experiment, a portion of stock solution was mixed with an equal volume of double-concentrated Tris-HCl buffer. Lipids. 1,2-Dimyristoyl-sn-glycero-3-phosphocholine (DMPC) and 1,2-distearoyl-sn-glycero-3-phosphocholine (DSPC) were purchased from Avanti Polar Lipids (Albaster, AL), and 1-palmitoyl-2-oleoyl-snglycero-3-phosphocholine (POPC) and 1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-(1′-rac-glycerol) (POPG) were obtained from NOF Corporation (Tokyo, Japan). Lipids were mixed with a 2:1 (v/v) chloroform−methanol mixture, and the solvent was removed by evaporation in a rotary evaporator. After drying under vacuum overnight, the residual lipid thin film was hydrated with buffer (10 mM Tris, 150 mM NaCl, 1 mM EDTA, pH 7.0, for PC or 10 mM Tris, 1 mM EDTA, pH 7.0, for PG) and vortex mixed to produce multilamellar vesicles. For the preparation of large unilamellar vesicles (LUVs), the suspension was subjected to seven cycles of freezing and thawing and then extruded through polycarbonate filters (100 nm pore size filter, 31 times) using a LiposoFast extruder (Avestin, Ottawa, Canada).14 Extrusion was carried out above the Tm of lipids. Small unilamellar vesicles (SUVs) were prepared by the sonication of multilamellar vesicles under a nitrogen atmosphere for 15 min (five cycles of 3 min) using a UD-200 probe-type sonicator (Tomy, Tokyo, Japan).17 Metal debris from the titanium tip of the probe was removed by centrifugation at 10 000g for 10 min. The diameters of the vesicles obtained were determined by dynamic light scattering using an FPAR1000 instrument (Otsuka Electronics, Osaka, Japan) and are summarized in Table 2. The lipid concentration of PC was determined

■

RESULTS Aβ Binding. We investigated the influence of the dimensions of the lipid vesicles and of the membrane curvatures on the interaction of Aβ[Y10W] with membranes using SUVs (diameter of ∼30 nm) and LUVs (diameter >100 nm) (Table 2). For our analysis, we chose membranes composed of electrically neutral PC (Figure 1) or anionic PG (Figure 2). It has been shown that Aβ interacts with electrostatically neutral lipid membranes in the gel phase and with anionic lipid membranes in the liquid-crystalline phase. By contrast, it does not bind neutral membranes in the liquidcrystalline phase.14,15,31 We used DMPC vesicles as neutral vesicles in the gel phase and POPC vesicles as neutral vesicles B

DOI: 10.1021/acs.langmuir.5b03332 Langmuir XXXX, XXX, XXX−XXX

Article

Langmuir

Figure 1. Tryptophan fluorescence of Aβs in the presence of PC vesicles at 4 or 25 °C. (a) Changes in fluorescence spectra as a function of the lipidto-protein ratio (L/P) as Aβ[Y10W] (2.5 μM) was titrated with DMPC-SUVs. (b) Changes in the spectra following the addition of DMPC-SUVs to Aβ[Y10W/H13K]. (c) Changes in the increment of the fluorescence intensity, R, as a function of L/P. Systems: Aβ[Y10W]-DMPC-SUVs (red squares), Aβ[Y10W]-DMPC-LUVs (black circles), Aβ[Y10W]-POPC-SUVs (blue inverted triangles), and Aβ[Y10W/H13K]-DMPC-SUVs (green triangles). (d) Changes in the maximum fluorescence wavelength, λmax. Symbols are identical to those in panel c.

Figure 2. Tryptophan fluorescence of Aβs in the presence of PG vesicles at 25 °C. (a) Changes in the fluorescence spectra as a function of the lipidto-protein ratio (L/P) as Aβ[Y10W] (2.5 μM) was titrated with POPG-SUVs. (b) Changes in the spectra following the addition of POPG-SUVs to Aβ[Y10W/H13K]. (c) Changes in the increment of the fluorescence intensity, R, as a function of L/P. Systems: Aβ[Y10W]-POPG-SUVs (red squares), Aβ[Y10W]-POPG-LUVs (black circles), and Aβ[Y10W/H13K]-POPG-SUVs (green triangles). (d) Changes in the maximum fluorescence wavelength, λmax. Symbols are identical to those in panel c.

in the liquid-crystalline phase at 4 °C. We also employed POPG vesicles as anionic vesicles in the liquid-crystalline phase at 25 °C. Aβ[Y10W] is useful for investigating the interaction between Aβ and lipids because the intensity of the tryptophan fluorescence increases and the λmax of the emission peak shows a blue shift when this residue is embedded in the hydrophobic environment of the lipid bilayers.32 In addition, it forms amyloid fibrils that are indistinguishable from Aβ-(1−40) amyloid fibrils.33 As a quantitative measure, the relative

fluorescence enhancement (R) is defined as R = (F − F0)/F0, where F and F0 denote the fluorescence intensities in the presence and absence of lipid membranes, respectively. For each system, changes in the fluorescence intensities were analyzed at the wavelength showing the maximum intensity at the largest L/P ratio. The addition of DMPC-SUVs caused variations in the tryptophan fluorescence intensity and in the position of the maximum wavelength, λmax (Figure 1a,c,d), whereas the addition of DMPC-LUVs or POPC-SUVs did not C

DOI: 10.1021/acs.langmuir.5b03332 Langmuir XXXX, XXX, XXX−XXX

Article

Langmuir ⎛ R ⎞ c f = ⎜1 − ⎟[Aβ] R max ⎠ ⎝

induce any changes (Figure S1). The addition of POPG-SUVs also induced changes in fluorescence (Figure 2a,c,d), whereas the addition of POPG-LUVs did not (Figure S2). Hence, Aβ[Y10W] preferably bound to DMPC-SUVs and POPGSUVs among the lipid vesicles examined. Interestingly, with the increase in the POPG-SUV concentration, the fluorescence intensity of Aβ[Y10W] decreased after reaching a maximum (Figure 2c). Such a reduction was not observed in the presence of DMPC-SUVs (Figure 1c) and suggested that Aβ bound to these systems with different modalities. Table 3 summarizes the binding parameters obtained by Langmuir’s equation:14,19

However, we could not estimate the binding parameters for the Aβ[Y10W]-POPG-SUVs system due to the decrease in R at L/ P ratios higher than 32. Therefore, we determined the binding parameters for Aβ[Y10W/H13K]-POPG-SUVs, where a monotonic increase in fluorescence intensity was observed (Figure 2c, see below). Secondary Structures. The conformational change in Aβ[Y10W] in the presence and absence of vesicles was investigated with CD measurements (Figures 3 and 4). The spectral variations observed when DMPC-SUVs and POPGSUVs were added to Aβ[Y10W] suggested that the peptide underwent a structural change in the presence of the two vesicles (Figures 3a and 4a). No spectral changes were observed when DMPC-LUVs, POPG-LUVs, and POPCSUVs were added (Figures 3b, 4b, and 3c, respectively). These results are consistent with the fluorescence binding experiments. In the presence of DMPC-SUVs, the spectra exhibited two minima at 208 and 222 nm indicative of α-helixrich structures attached to SUVs (Figure 3a). In the presence of POPG-SUVs, Aβ formed β-strand-rich structures at low L/P ratios (≤20), as judged from the minimum at 218 nm, and αhelical structures at L/P ratios higher than 64 (Figure 4a). We hypothesized that the decrease in fluorescence intensities at high L/P ratios in the mixtures containing POPG-SUVs (Figure 2) is caused by the structural rearrangement experienced by the peptide. Upon transition from the β-strand to α-helix, the imidazole ring of histidine 13 would move closer to tryptophan 10 and quench its fluorescence.34 Assuming that the peptide adopted canonical structures, the interatomic distance between the Cγ atom of W10 and that of H13 would decrease from ∼1.2 nm to ∼0.7 nm. The latter distance is approximately equal to the one between histidine 18 and tryptophan 94 of barnase (0.6 nm), which shows quenching of the tryptophan fluorescence by the histidine residue.35 To elucidate the effect of H13, we performed similar titration experiments using Aβ[Y10W/

Table 3. Binding Parameters

a

system

K (× 107 M−1)

xmax (mol/mol)

Rmaxa

Aβ[Y10W]-DMPC Aβ[Y10W, H13K]-DMPC Aβ[Y10W, H13K]-POPG

1.75 ± 0.63 4.17 ± 1.00 1.72 ± 1.43

0.011 ± 0.001 0.016 ± 0.001 0.119 ± 0.010

1.55 1.25 0.32

Determined by extrapolation of the R vs 1/(L/P) plot to 0.

x=

xmaxKc f 1 + Kc f

Here, x and cf represent the ratio between the molar amount of the membrane-bound Aβ per exofacial lipid and the concentration of free Aβ in solution, respectively. K and xmax denote the association constant and the maximal amount of membrane-bound Aβ per lipid, respectively. The fraction of membrane-bound Aβ is given by R/Rmax, where Rmax is the maximal fluorescence enhancement in Figures 1c and 2c. Then, x and cf are obtained as follows: x=

2R[Aβ] R max[lipid]

Figure 3. CD spectra of Aβs in the presence of PC vesicles at 4 or 25 °C. (a) Changes in spectra of Aβ[Y10W] (2.5 μM) upon addition of DMPCSUVs. (b) Spectra of Aβ[Y10W] in the presence of DMPC-LUVs. (c) Spectra of Aβ[Y10W] in the presence of POPC-SUVs. (d) Spectra of wildtype Aβ in the presence of DMPC-SUVs. L/P ratios: 0 (black), 64 (red), 120 (yellow), 240 (green), and 640 (cyan). D

DOI: 10.1021/acs.langmuir.5b03332 Langmuir XXXX, XXX, XXX−XXX

Article

Langmuir

Figure 4. CD spectra of Aβs in the presence of PG vesicles at 4 °C. (a) Changes in spectra of Aβ[Y10W] (2.5 μM) upon addition of POPG-SUVs. (b) Spectra of Aβ[Y10W] in the presence of POPG-LUVs. (c) Spectra of Aβ[Y10W/H13K] in the presence of POPG-SUVs. (d) Spectra of wildtype Aβ in the presence of POPG-SUVs. L/P ratios: 0 (black), 12 (red), 20 (yellow), 64 (green), and 96 (cyan).

Figure 5. Fibrillation of wild-type Aβ in the presence of SUVs at 37 °C. (a−c) Increase in ThT fluorescence intensity as a function of time in the presence of DSPC-SUVs (a), POPG-SUVs (b), and POPC-SUVs (c) at L/P ratios of 0 (black circles), 4 (red squares), 20 (blue triangles), 50 (green inverted triangles), and 200 (purple diamonds). (d, e) TEM images of Aβ incubated for 7 days with DSPC-SUVs (d) and POPG-SUVs (e).

H13K] and monitored the fluorescence and CD signals in the presence of POPG-SUVs. In contrast to Aβ[Y10W], the R value of Aβ[Y10W/H13K] did not decline when the L/P ratio exceeded 20 (Figure 2b,c), while the CD spectra of Aβ[Y10W/ H13K] represented similar structural transitions to those of Aβ[Y10W] (Figure 4a,c). These results supported our hypothesis that the tryptophan fluorescence of Aβ[Y10W] was quenched by histidine 13 when the peptide adopted an αhelical structure in the presence of POPG-SUVs. On the other hand, the fluorescence variation of Aβ[Y10W/H13K] was similar to that of Aβ[Y10W] in the presence of DMPC-SUVs (Figure 1b−d), although a slight difference in the binding

parameters was observed (Table 3). This result indicated that the region containing the W10 and H13 residues did not form an α-helix when Aβ formed α-helix-rich structures on the surface of DMPC-SUVs. Therefore, we conclude that the αhelical conformations forming on POPG-SUVs and on DMPCSUVs are significantly different. We also performed CD measurements for wild-type Aβ-(1−40) (Figures 3d and 4d). Upon addition of vesicles (DMPC- or POPG-SUVs), the spectra of wild-type Aβ changed in a similar manner to those of Aβ[Y10W], suggesting that wild-type Aβ and Aβ[Y10W] bind to lipid membranes with the same mechanism. Taken together, the above observations demonstrate that wild-type Aβ and the E

DOI: 10.1021/acs.langmuir.5b03332 Langmuir XXXX, XXX, XXX−XXX

Article

Langmuir

due to the fact that they carried out their study at pH 6, where the interaction between positively charged Aβ and negatively charged LUVs has a mainly electrostatic nature, independent of the vesicle curvature. A number of studies have demonstrated that smaller vesicles facilitate the surface binding of amphiphilic peptides or proteins, such as ALPS motifs and BAR domains known as curvature-sensing proteins.39 Amyloidogenic proteins, such as polyglutamine domains and α-synuclein, also display curvaturedependent binding and aggregation on vesicle surfaces.40,41 A molecular dynamics study also showed that transmembrane structures of the C99 fragment of amyloid precursor protein were influenced by the membrane surface curvature.42 Packing defects of lipid bilayers might play key roles in interaction in these membrane−protein systems. Bending of the membranes modifies the packing density of the lipid molecules, exposing the hydrophobic core and reducing lateral pressure. Note that packing defects in liquid-crystalline lipids appear to be insufficient for binding Aβ to the surface, as demonstrated by the fact that POPC-SUVs exhibit no interactions with the peptide. This observation is consistent with the study of molecular dynamics simulation, where Aβ exhibited only shortlived contacts with a POPC bilayer.43 Electrostatic interactions with charged lipids such as POPG or large packing defects in the gel phase as described below are likely required for binding. We determined the binding parameters between Aβ and SUVs using the Langmuir adsorption model (Table 3). Here, we assume the thermodynamic equilibrium between the nonaggregated protein in solution and the bound one on membranes. Although subsequent aggregation can occur after prolonged incubation (Figure 5), we confirmed that the aggregation kinetics of Aβ was sufficiently slower than the Aβ-membrane binding. In addition, freshly prepared Aβ had random coil structure in buffer as judged from the CD spectrum (Figures 3 and 4) and exhibited no ThT fluorescence signals (Figure 5), suggesting the nonaggregated monomeric protein. DMPC-SUVs and POPG-SUVs display a different number of binding sites for Aβ, xmax, while the binding constants, K, are almost identical. The xmax for gel-phase DMPC-SUVs is 10-fold smaller than that for POPG-SUVs. Cryo-TEM images of small vesicles in the gel phase showed that they form polyhedral shapes with many small planar facets jointed by quasi-spherical edges.44 Isothermal calorimetry indicated that bending of the membranes induced an increase in their enthalpy.45 It is possible that hydrophobic acyl chains are more exposed to water if the membranes edges are bent, forming a site for the deep insertion of Aβ in the membranes (Figure S4). This is supported by the maximal λmax at 337 nm in the tryptophan fluorescence spectra in the presence of DMPC-SUVs (Figure 1), which indicates that the tryptophan residue of Aβ is located deeply in the hydrophobic region at a distance of ∼7.5 Å from the bilayer center.46 The limited region of the surface accessible to Aβ does not conflict with the number of binding sites, xmax, of 0.011. The surface area for one Aβ molecule (∼9 nm2) and the area of a binding site formed by the lipids in the outer leaflet of the membrane (0.6 nm2 per lipid × 1/xmax ≈ 54 nm2) indicate that only one-sixth of the outer lipids provides the binding sites for Aβ. This value is consistent with a ratio between the areas of the planar facets and the quasi-spherical edges of ∼5:1, roughly estimated assuming that the SUV is a regular dodecahedron with round edges (with a circumsphere radius of 20 nm and a rounding radius of 1 nm). Note that we assume that only headgroups of

mutants used in this study interact with the vesicles in the same manner. Fibrillation. The formation of amyloid fibrils was monitored by the increase in fluorescence of amyloid-specific dye ThT (Figure 5a−c). The ThT fluorescence intensity remained constant for 14 days when Aβ was incubated in the absence of SUVs or in the presence of POPC-SUVs at L/P ratios of 4−50 at 37 °C, suggesting that the peptide did not form amyloid fibrils. A slight increase in the ThT fluorescence was observed in POPC-SUVs at an L/P ratio of 200. Fibrillation of wild-type Aβ was not observed in the presence of DMPC-SUVs (Tm: 24 °C) incubated at 4 °C for 14 days (data not shown). To monitor the aggregation of Aβ at 37 °C in the presence of SUVs that are in the gel phase, we used DSPC (Tm: 55 °C) instead of DMPC. The ThT fluorescence intensity increased in the presence of DSPC-SUVs as a function of time, indicating the formation of fibrils. At an L/P ratio of 50, the aggregation was much faster than at L/P ratios of 4, 20, and 200. Similar to DSPC-SUVs, a substantial increase in Aβ aggregation was observed in the presence of POPG-SUVs at an L/P ratio of 4 but not at L/P ratios of 20−200 at 37 °C. In addition, we observed no lag times in aggregation. We did not observe Aβ aggregation in the presence of POPG-SUVs at 4 °C or in the presence DMPC-SUVs, while the tryptophan fluorescence spectra showed Aβ-POPG interactions (Figure S3), probably because the low temperature prevents the formation of amyloid fibrils. These results suggested that the high Aβ densities deriving from almost complete occupation of the available binding sites on the DSPC- and POPG-SUVs triggered the nucleation and formation of amyloid fibrils at 37 °C. The formation of amyloid fibrils was also confirmed by TEM (Figure 5d,e).

■

DISCUSSION Yoda et al. showed that Aβ-(1−40) binds to PC vesicles when they are in the gel phase but not when they are in the liquidcrystalline phase.15 They speculated that a flat surface of tightly packed lipids serves as a binding site for the protein through nonelectrostatic interactions. In the present study, we used tryptophan fluorescence and CD studies to confirm that Aβ binds to DMPC-SUVs in the gel phase to form α-helix-rich structures and that it does not bind to POPC-SUVs in the liquid-crystalline phase. These observations are consistent with a previous report that detected a 3 nm blue shift of tryptophan fluorescence and a weak change in the CD spectrum of Aβ in the presence of 1,2-dipalmitoyl-PC SUVs at L/P ratios of