Fusogenic properties of the C-terminal domain of the Alzheimer beta-amyloid peptide.

A series of natural peptides and mutants, derived from the Alzheimer β-amyloid peptide, was synthesized, and the potential of these peptides to induce fusion of unilamellar lipid vesicles was investigated. These peptide domains were identified by computer modeling and correspond to respectively the C-terminal (e.g. residues 29-40 and 29-42) and a central domain (13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28) of the β-amyloid peptide. The C-terminal peptides are predicted to insert in an oblique way into a lipid membrane through their N-terminal end, while the mutants are either parallel or perpendicular to the lipid bilayer. Peptide-induced vesicle fusion was demonstrated by several techniques, including lipid-mixing and core-mixing assays using pyrene-labeled vesicles. The effect of peptide elongation toward the N-terminal end of the entire β-amyloid peptide was also investigated. Peptides corresponding to residues 22-42 and 12-42 were tested using the same techniques. Both the 29-40 and 29-42 β-amyloid peptides were able to induce fusion of unilamellar lipid vesicles and calcein leakage, and the amyloid 29-42 peptide was the most potent fusogenic peptide. Neither the two mutants or the 13-28 β-amyloid peptide had any fusogenic activity. Circular dichroism measurements showed an increase of the α-helical content of the two C-terminal peptides at increasing concentrations of trifluoroethanol, which was accompanied by an increase of the fusogenic potential of the peptides. Our data suggest that the α-helical content and the angle of insertion of the peptide into a lipid bilayer are critical for the fusogenic activity of the C-terminal domain of the amyloid peptide. The differences observed between the fusogenic capacity of the amyloid 29-40 and 29-42 peptides might result from differences in the degree of penetration of the peptides into the membrane and the resulting membrane destabilization. The longer peptides, residues 22-42 and 12-42, had decreased, but significant, fusogenic properties associated with perturbation of the membrane permeability. These data suggest that the fusogenic properties of the C-terminal domain of the β-amyloid peptide might contribute to the cytotoxicity of the peptide by destabilizing the cell membrane.

The amyloid peptide (A␤), 1 a 39 -43-residue peptide, is a normal 4-kDa derivative of a large transmembrane glycoprotein, the amyloid ␤ precursor protein. The A␤ peptide is found in an aggregated, poorly soluble form in extracellular amyloid deposition in the brains and leptomeniges of patients with Alzheimer's disease (1). In contrast, it occurs in a soluble form in several biological fluids, including the cerebrospinal fluid, where it is produced by glial cells and neurons and where it circulates at nanomolar concentrations (2). The mechanism by which A␤ causes cell death and exerts its cytotoxicity effect remains unclear, and controversies still exist concerning the cytotoxic action of A␤ on neuronal cells. A number of in vitro studies with the synthetic A␤ peptide have shown that this peptide aggregates easily and forms amyloid fibrils similar to that found in the brain of patients with Alzheimer's disease (3)(4)(5). Recent observations have demonstrated that the C-terminal domain (amino acids 34 -42) of A␤ is critical for amyloid aggregation and fibril formation (6).
Aggregated A␤ has been shown to be cytotoxic for cultured neurones, whereas nonaggregated peptide showed less cytotoxic effects (7)(8)(9). Arispe et al. (10,11) have found that amyloid peptides were able to form calcium channels in bilayer membranes and other studies have suggested that the amyloid 1-40 and 1-42 peptides increased the intraneuronal free calcium concentration (12,13). Moreover, recent reports have demonstrated a cytotoxic effect of A␤, at relatively high doses (10 -40 M) on neuronal cells via an oxidative mechanism, which was not mediated by a receptor pathway (14,15). Altogether, these data suggest a direct perturbation of cell membranes by the amyloid peptide. The mode of interaction of the peptide with the membrane has not been elucidated yet, and it is not clear whether aggregated A␤ can directly affect the cell membrane permeability or whether it enhances the membrane susceptibility to further injuries.
Membrane destabilization, due to the tilted penetration of fusion peptides into cellular membranes, is induced by a variety of peptides (16,17). These hydrophobic peptides penetrate the lipid bilayer at an angle of 45-50°, due to the hydrophobicity gradient along their sequence (18,19). Computer modeling of the ␤-amyloid peptide showed that the C-terminal domain of A␤ (e.g. amino acids 29 -40 or 29 -42) has properties similar to those of the fusion peptide of viral proteins (20). We therefore investigated the fusogenic properties of the C-terminal domain of amyloid peptide and report in this paper, that both the amyloid 29 -40 and amyloid 29 -42 peptides are able to cause fusion of unilamellar lipid vesicles with concomitant perturbation of membrane permeability as monitored by calcein leakage and that the amyloid 29 -42 peptide is the most fusogenic. The ␣-helical structure and the obliquity of the peptides seem critical for their interaction with a lipid bilayer and for their fusogenic activity. The longer amyloid peptide, A␤-  and A␤- , had decreased but significative fusogenic activity, also associated with calcein release from the vesicles. This suggests that penetration of residues 29 -42 into the membrane is hampered by N-terminal residues. Our results propose an alternative mechanism for the amyloid peptide cytotoxicity through a direct perturbation of the cellular plasma membrane similar to that induced by viral fusion peptides. The penetration of the cell membrane by the peptide at a tilted angle causes membrane perturbation and further cell damages.
Synthesis and Purification of Peptides-Peptides were synthesized by the standard Fmoc (N-(9-fluorenyl)methoxycarbonyl) solid-phase method, on an Applied Biosystems model 431A peptide synthesizer (21). The peptide-resin conjugate was cleaved with trifluoroacetic acid, and the peptide was precipitated with tributylmethyl ether and recovered by centrifugation at 2000 ϫ g. The residue was dried for 2 h in a Speedvac concentrator (Savant Instruments, Farmingdale, NY). The peptides (20 -40 mg) were resuspended in acetonitril/water (80/10, v/v) to get rid of all scavengers molecules and the insoluble peptide was recovered by centrifugation. The washing step with water was repeated 10 times. The purity and correct sequences of all peptides were verified by electron spray ionization mass spectrometry using a Fisons/VG Platform (Manchester, United Kingdom) mass. Molecular Modeling of Peptides by Energy Minimization-Modelization of all peptides was carried out as described previously (16), and the method used is that applied to the study of polypeptide conformation (22). The method used to predict conformational structure of the peptides accounted for the contribution of the lipid-water interface, the concomitant variation of the dielectric constant, and the transfer energy of atoms from a hydrophobic to hydrophilic environment (23). The structure, mode of insertion, and orientation of the peptides were studied in a dipalmitoylphosphatidylcholine monolayer. In this model, the interaction energy (sum of contributions from Van der Waals energy interactions, torsional potential energy, electrostatic interactions, and transfer energy between peptide and dipalmitoylphosphatidylcholine in the monolayer) was calculated and minimized until the lowest energy state of the entire aggregate was reached. All calculations were performed on an Olivetti CP486, by using PC-TAMMOϩ (Theoretical Analysis of Molecular Membrane Organization) and PC-PROTϩ (Protein Plus Analysis) software. Graphs were drawn with the PC-MGMϩ (Molecular Graphics Manipulation) program.
Circular Dichroism Measurements-To overcome problems of peptide solubility at high concentration, fresh peptide stock solutions (25 mg/ml) were prepared in hexafluoro-2-propanol, and peptides were further diluted in TFE. Circular dichroism spectra were obtained at 23°C in a Jasco 710 spectropolarimeter calibrated with 0.1% (w/v) D-10-camphorsulfonic acid solution (24). Portions of peptide stock solutions were removed and diluted in a 10 mM sodium phosphate buffer, pH 7.4, containing different percentages of TFE. Nine spectra were recorded 20 min after the peptides were dissolved in the aqueous buffer/TFE solution at a final concentration of 0.1 mg/ml. The percentage of secondary structure were estimated by curve fitting on the entire ellipticity curve between 184 and 260 nm according to the variable selection procedure developed by Johnson (25).
Lipid Mixing Experiments-Small unilamellar vesicles (SUVs) were prepared from a mixture of PC/PE/PS/SM/cholesterol (10:5:7.5:7.5:16, w/w), when necessary in the presence of 2.5 mol% of Pyr-PC. The composition of the vesicles was selected to mimick that of rat neuronal membranes (26). All lipids were dissolved in chloroform at 10 mg/ml and dried under a stream of nitrogen. Pyr-PC was dissolved in chloroform/ethanol at a concentration of 2 M. Dried mixed lipids were hydrated in a 10 mM Tris-HCl buffer, pH 8.0, containing 150 mM NaCl, 0.01% (w/v) Na-EDTA, and 1 mM NaN 3 . The lipid suspension was sonicated at 23°C, using a Branson sonifier, under nitrogen at 32 watts for 4 ϫ 15 min. After sonication, the labeled and unlabeled vesicles were applied to a Sepharose CL-4B column. SUVs were separated from larger particles eluting in the void volume of the column. In order to obtain an homogenous preparation, only the top fractions of the SUV elution peak were collected and pooled. Phospholipid concentration was determined by an enzymatic assay (BioMérieux, Charbonnier les Bains, France). Large unilamellar vesicles (LUVs) were prepared from phospholipids as follows. Dry lipids were hydrated in buffer and dispersed by vortexing to produce large multilamellar vesicles. The lipid suspension was freeze-thawed five times and then extruded 10 times through two stacked 100-nm pore size polycarbonate filters in a pressure extruder (Lipex Biomembranes, Inc., Vancouver, Canada). The LUV and SUV preparations could be stored at 4°C for at least 1 week, without significant size change of the vesicles during storage. Fusion of pyrenelabeled SUVs together with unlabeled vesicles, at a 1/4, w/w, ratio, was measured using a fluorescence probe dilution assay. The pyrene excimer/monomer intensity ratio was measured as a function of time after addition of increasing quantities of the peptides dissolved at a concentration of 1 mg/ml in either 50 or 20% (v/v) TFE. Fusion resulted in a decrease of the excimer intensity and a slight increase of the monomer fluorescence. Emission spectra were obtained on an Aminco SPF 500 spectrofluorimeter at 25°C. The pyrene excimer/monomer (E/M) ratio was calculated from the excimer and monomer fluorescence intensity at 475 and 398 nm, respectively, with an excitation wavelength of 346 nm. The excitation and emission bandwidth were set at 2 and 5 nm, respectively. Under these conditions, unlabeled vesicles in the presence of the peptides gave no significant signal due to light scattering. All experiments were performed in a 10 mM Tris-HCl buffer, pH 8.0, containing 150 mM NaCl, and 0.1 g/liter Na-EDTA for a final volume of 500 l. The final TFE concentration in the reaction mixture was less than 5%, and a correction was made for the effect of the solvent on the excimer/ monomer ratio. The final peptide concentrations were in the range of 6.9 to 46 M.
Visible Absorbance Measurements-The changes in the vesicle size distribution were monitored by visible absorbance measurements, as described previously (27). Aliquots of peptide stock solution (1 mg/ml in 50% TFE) were added to 15 g of SUVs as for the fusion monitoring by fluorescence measurement. The absorbance at 405 nm was monitored in a function of time on a Uvikon 940 spectrophotometer.
Monitoring of Vesicle Size Increase by Gel Filtration-The ability of the labeled and unlabeled SUVs to undergo fusion was assessed by gel filtration on a Sepharose CL-4B column before and after the addition of the peptides. During gel filtration, the fluorescence emission intensity of Pyr-PC was continuously monitored.
Core-mixing Experiments-Vesicle fusion was also investigated by a core-mixing assay developed by Kendall and MacDonald (28). Lipid vesicles were prepared as described above except that the phospholipid mixture was prepared by hydrating 2 mg of phospholipids in 1.5 ml of a 10 mM Tris-HCl buffer, pH 8.0, containing 150 mM NaCl and 1 mM NaN 3 and containing calcein at 0.8 mM plus CoCl 2 at 1.0 mM or EDTA at 20 mM. Vesicles were then vigorously vortexed for 2 min and sonicated as described previously. Untrapped solutes were removed by two successive elutions on a Sephadex G-25 column with 10 mM Tris-HCl buffer, pH 8.0, containing 150 mM NaCl, and 1 mM NaN 3 . The lipid concentrations of the liposome suspensions were determined by phosphorus analysis (29). In a standard experiment, calcein, Co 2ϩ -and EDTA-containing vesicles were mixed at a 1:1 molar ratio in a 10 mM Tris-HCl buffer, pH 8.0, containing 150 mM NaCl and 1 mM NaN 3 . Peptides from concentrated stock solutions were added at different concentrations and the calcein fluorescence was followed using an Aminco SPF 500 spectrofluorimeter, with excitation and emission wavelengths of 490 and 520 nm, respectively, and with excitation and emission slits of 5 nm. The maximum fluorescence yield was determined in the presence of 0.5% Triton X-100 and EDTA 10 mM. In some experiments, CoCl 2 at 0.4 mM chelated with 0.4 mM citrate were present in the reaction mixture to assess the leakage of encapsulated components. All experiments were performed under constant stirring at 25°C in the fluorimeter cuvette.
Leakage of Liposome Contents-The leakage of liposome contents to the external medium was kinetically monitored by measuring the re-lease of calcein trapped inside the vesicles (30). The SUVs were prepared as described previously except that dried lipids were rehydrated in a 10 mM Tris-HCl buffer, pH 8.0, containing 150 mM NaCl, 0.1 g/liter Na-EDTA, 1 mM NaN 3 , and 40 mM calcein. At this concentration, fluorescence self-quenching occurs. Nontrapped calcein was removed by gel chromatography on Sephadex G-25. Vesicles were eluted with a 10 mM Tris-HCl buffer, pH 8.0, containing 150 mM NaCl, 0.1 g/liter Na-EDTA and 1 mM NaN 3 . The lipid concentrations of the liposome suspensions were determined by phosphorus analysis (29). Calcein entrapped at a self-quenching concentration in SUVs increases in fluorescence when it leaks from inside the liposomes. Upon addition of the peptides from a 50% (v/v) TFE stock solution, the increase in calcein fluorescence was followed, using an Aminco SPF 500 spectrofluorimeter, with excitation and emission wavelengths of 490 and 520 nm, respectively, and with excitation and emission slits of 5 nm. 100% leakage was established by lysis of the vesicles with 0.5% (w/v) Triton X-100 (30). The percentage of calcein release is defined as: % F(t) ϭ (I t Ϫ I 0 /I tot Ϫ I 0 ) ϫ 100, where I 0 ϭ initial fluorescence, I tot ϭ total fluorescence observed after addition of Triton X-100, and I t ϭ fluorescence at time t. In our experimental conditions, in the absence of peptide, the spontaneous leakage rate was less than 1%/10 min. The volume of the reaction mixture was fixed at 500 l. All experiments were with constant stirring at 25°C.

RESULTS
The sequences and properties of the synthetic peptides representing different domains of the Alzheimer amyloid peptide (A␤) are summarized in Table I. The central domain of the peptide A␤ is represented by the A␤-(13-28) peptide, while peptides of the C-terminal domain of the A␤ peptide are designed as A␤- (29 -40) and A␤-(29 -42), respectively. In addition, two mutants of the A ␤-(29 -42) peptide, named A␤-(29 -42, 0°) and A␤-(29 -42, 85°), were synthesized. These peptides have the same amino acid composition as the wild-type peptide, but their sequence was designed in order to change the angle of insertion of the peptides in a lipid bilayer. They are oriented at an angle of, respectively, 0 and 85°at a lipid/water interface, compared with an angle of 50°for the wild-type peptides (Table  I). This obliquity is shared by viral fusion peptides, such as the Simian immunodeficiency virus fusogenic peptide (SIV) listed in Table I, whose properties have been extensively described (18). The amyloid C-terminal peptides are all highly hydrophobic, with an hydrophobicity close to that of a transmembrane peptide on the Eisenberg's scale (16). Compared with the Cterminal peptides, the 13-28 central peptide is much less hydrophobic and is not predicted to form a stable helix.
Computer Modeling of the Amyloid Peptides-In analogy with the calculations performed on the viral fusogenic peptides (20), the conformation of the ␤-amyloid peptides at a lipid/ water interface was calculated by energy minimization (16). The most probable conformation of the peptides is illustrated in Fig. 1, showing the oblique orientation of the A␤-(29 -40) and A␤-(29 -42) peptides and of the SIV fusion peptide and the different orientation of the mutants. The oblique orientation of the wild-type peptides compared with the lipid bilayer is due to the concentration of the most hydrophobic residues at the Nterminal extremity of the peptide, thereby creating a N-C hydrophobicity gradient along the sequence. As described above, the peptides penetrate the phospholipid bilayer or insert into the lipid phase through their more hydrophobic N-terminal end and the 7-8 N-terminal residues insert into the lipid to destabilize the regular packing of the phospholipid acyl chains (Fig.  1).
Lipid Mixing Induced by the Tilted Peptides-The induction of intervesicular lipid mixing by peptides, as a measure of their fusogenic activity, was tested with PC/PE/PS/SPH/cholesterol SUVs utilizing a probe dilution assay (27,31). SUVs labeled with pyrene-lecithin were mixed with unlabeled vesicles, and the ratio of the excimer to monomer intensity of the pyrene probe was measured as a function of time. A decrease of the excimer intensity and an increase of the monomer intensity due to the dilution of the probe into the fused vesicles is a measure of the fusion. As shown in Fig. 3, among the six peptides tested, only A␤- (29 -40) and A␤-(29 -42) had fusogenic properties monitored by the decrease of the E/M ratio of mixed unlabeled and labeled SUVs. The A␤-(13-28), A␤-(29 -42, 0°), and A␤-(29 -42, 85°) peptides had not fusogenic activity, as predicted from computer modeling and from the calculation of the peptide obliquity (Table I). In a control experiment, the SIV peptide, with structural characteristics similar to those of the C-terminal domain of the amyloid peptide, showed fusogenic activity, thus confirming the validity of the experimental approach (Fig. 3). Similar results were obtained using labeled and unlabeled LUV preparations to monitor lipid mixing induced by the amyloid and the SIV peptides (data not shown). The dependence of both the extent and kinetics of the lipid mixing process, on the peptides concentration, was examined. In separate experiments, increasing amounts of A␤-(29 -40) and A␤-(29 -42) peptides were added to a fixed amount of SUVs. The time course variation of the E/M ratio induced by adding various amounts of the two peptides showed that both peptides are able to cause lipid mixing, in a concentration-dependent manner. To compare the activity of the two peptides, the level of the E/M ratio decrease 10 min after the addition of the peptide was measured and is depicted on Fig. 4. A␤-(29 -42) is more active in inducing SUVs fusion than the A␤-(29 -40) peptide, as the A␤-(29 -42) peptide causes fusion more rapidly and to a higher extent than A␤- (29 -40). At 46 M, the A␤-(29 -42) caused a 71.5% of E/M ratio decrease after 10 min of incubation with the vesicles, whereas at the same peptide concentration, the E/M ratio decreased of 59% when the A␤-(29 -40) peptide was mixed with the vesicles. Same results were obtained using PC/PE/cholesterol LUVs (data not shown). In control experiments, incubation of mixed labeled and unlabeled vesicles, in the absence of peptide or incubation of labeled vesicles alone with the A␤- (29 -40) or A␤-(29 -42) peptides, had no effect on the E/M ratio (data not shown). These results strongly suggest that the observed lipid mixing induced by the amyloid peptides is the result of intervesicular lipid mixing, as suggested by Morris et al. (31). In order to assess the influence of the secondary structure of the peptides on their fusogenic activity, the peptides were diluted in 20 or 50% TFE and added to the SUV preparation (Fig. 5). Compared with the effect measured with peptides dissolved in 50% TFE, solubilizing the peptides in 20% TFE decreased intervesicular lipid mixing. As the percentage ␣-helical structure of the peptides is maximal in 50% TFE, as shown in Fig. 2, this suggests that the fusogenic properties of the peptides increase with their helical conformation. This effect is more pronounced for the A␤-(29 -42) peptide whose ␣-helical content increases most between 20 and 50% TFE.
Visible Absorbance Studies and Gel Filtration-Changes in vesicles size distribution resulting from fusion can be monitored by following the absorbance of the SUV preparation (27). The changes of absorbance at 405 nm as a function of time after addition of the peptides to SUV suspension are plotted in Fig.  6, A and B. The data show that, at all concentrations tested, A␤-(29 -40) and A␤-(29 -42) induce an increase of vesicle size. The initial rate and the total amplitude of the absorbance increase are highest for the A␤-(29 -42) peptide, and these differences are more pronounced at higher peptide concentrations (46 M). At this peptide concentration, the absorbance measured after addition of A␤-(29 -42) reached its final level within 4 -5 min, whereas the A␤-(29 -40) peptide initially induced a moderate increase of absorbance followed by a slower increase. These data are in agreement with those obtained by fluorescence measurements (Figs. 3 and 4). The SIV fusion peptide, at a concentration of 13.8 M, had a similar effect on vesicular size, whereas the A␤-(29 -42, 0°) and A␤-(29 -42, 80°) peptides had no effect (Fig. 6C). To confirm that the turbidity increase and the intervesicular lipid mixing resulted of membrane fusion, the increase of vesicle size induced by the amyloid peptides was also monitored by gel filtration on a Sepharose CL-4B column. Fig. 7 shows that incubation of an homogenous population of pyrene-labeled vesicles with unlabeled vesicles, in the presence of the A␤-(29 -40) and A␤-(29 -42) peptides, significantly increases the size of the particles. The pyrene intensity is maximum within the void volume of the Sepharose CL-4B column. By comparison, the mixture eluted as a major peak within the separation volume of the column when no peptide or the A␤-(13-28) peptide was added (Fig. 7).

Lipid Mixing Induced by the A␤-(22-42) and A␤-(12-42)
Peptides-Computer modelization of the amyloid peptide carried out on the C-terminal domain of the ␤-amyloid peptide predicted that it inserts into the lipid bilayer via its N-terminal residues (Fig. 1). In order to investigate the effect of the elongation of the peptide at its N-terminal end, we synthesized longer peptides spanning residues 22-42 and 12-42, A␤-(22-42) and A␤-(12-42) peptides, respectively. As shown in Fig. 8, both peptides were able still to induce fusion of SUVs measured by lipid-mixing assay within the same concentration range as described above. The same results were obtained using labeled and unlabeled LUVs (data not shown). When the peptides were incubated at 46 M with a mixture of labeled and unlabeled SUVs, the A␤-(22-42) peptide induced a E/M ratio decrease of about 48% and the A␤-(12-42) peptide a decrease of 37%, compared with 72% for the A␤-(29 -42) peptide. Incubation of the peptides with pyrene-labeled vesicles in the absence of unlabeled vesicles showed no decrease of the pyrene excimer/ monomer ratio, suggesting that the observed fusion was due to membrane intermixing and was not due to exchange of pyrenelabeled phospholipids facilitated by simple vesicle aggregation (data not shown). These results suggest that the addition of N-terminal residues of the A␤ peptide has only limited effects on the interaction of the C-terminal fusogenic peptide with a lipid bilayer and that this tilted peptide is still able to induce fusion within the longer A␤ peptides.
Core Mixing Induced by the Synthetic Amyloid Peptides-In order to further document the fusogenic properties of the amyloid peptides previously monitored by the lipid-mixing assay and supported by visible absorbance measurements and gel filtration, experiments were carried out to demonstrate contents mixing of two vesicle populations labeled separately (28). When the amyloid peptides were added to a mixture of calcein and Co 2ϩ -and EDTA-containing PC/PE/PS/SPH/cholesterol SUVs, an increase of the fluorescence was observed (data not shown). The observed increase in fluorescence due to increase concentration of free calcein, could be a consequence of EDTA-Co 2ϩ complex formation occurring within the inner compartment of the fused vesicles. On the other hand, if leakage of the vesicle contents occurs, dilution and subsequent dissociation of calcein-Co 2ϩ complex might take place. This would increase the emission intensity of the free calcein. Fluorescence due to leakage of vesicle contents can be eliminated by adding 0.4 mM Co 2ϩ (chelated in a 1:1 molar ratio with citrate) in the outer phase. Fig. 9 illustrates the effect of increasing concentrations of the different amyloid peptides on the core mixing of calcein and Co 2ϩ -and of EDTA-containing SUVs after 10 min of incubation in the presence of Co 2ϩ in the external phase. The results show that both peptides can induce content mixing of the two populations of vesicles within the same concentration range as in the lipid mixing experiments. In agreement with the lipidmixing assay, the A␤-(29 -42) peptide is the most fusogenic peptide as it induces a 23% increase of the calcein emission intensity at 46 M, compared with 19% for the A␤-(29 -40) peptide. As described above, the longer peptides are still able to induced vesicle fusion monitored by a core-mixing assay. In control experiments, we demonstrated that neither the A␤-(29 -42, 0°), the A␤-(29 -42, 85°), or the A␤-(12-28) peptides were able to cause significant content mixing of the two populations of vesicles (data not shown).
Peptide-induced Leakage of Liposomal Contents-Experi-ments from core-mixing assay suggested that concomitantly with membrane fusion, release of vesicle contents occurred in the presence of the amyloid peptides. To assess the leakage of vesicle contents due to the interaction with the synthetic amyloid peptide, we performed leakage measurement of encapsulated free calcein from SUVs. The addition of the C-terminal amyloid peptides to SUVs (15 g of phospholipids) produced membrane destabilization as measured by the release of calcein from the vesicles. Fig. 10 illustrates the concentration-dependent release of calcein from SUVs, induced by the addition of the A␤- (29 -40), A␤-(29 -42), A␤- , and A␤-  peptides. Within this concentration range, the rate of calcein leakage was significant for all peptides. As shown in Fig. 10, after 10 min of incubation, the A␤-(29 -42) peptide was more active than the A␤-(29 -40) peptide, in agreement with the results from the lipid-mixing assay (Fig. 4) and consistent with the previously discussed differences in bilayer penetration (Fig.  1). In agreement with the results of the lipid-and core-mixing assays, the A␤-(22-42) and the A␤-(11-42) peptides were also able to perturb the membrane of the vesicles and to induce calcein release from the SUVs. In control experiments, we observed that the A␤-(13-28) peptide caused a nonsignificant calcein release of 9.5 and 10.6% after 10 min of incubation at 23 and 46 M, respectively (data not shown). Under the same experimental conditions, the A␤-(29 -42, 0°) and the A␤-(29 -42, 85°) peptides at 46 M induced a calcein release of 19 and 23%, respectively (data not shown). We cannot rule out an aspecific interaction of these peptides with the SUVs due to their high hydrophobicity. This suggests however that the  tilted insertion of the peptides into the lipid enhanced the release of the encapsulated calcein. The limited effect of the A␤- (13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28) peptide and of the A␤-(29 -42) mutant peptides suggests that the leakage of the liposomal content induced by the wild-type peptide, is specific. Altogether, these results suggest that the fusion of SUVs induced by the amyloid peptides is correlated to lipid bilayer destabilization via a direct interaction of the amyloid peptide with the lipid phase and that this process is mediated by the C-terminal end of the peptide (e.g. amino acids 29 -42).

DISCUSSION
Several papers recently appeared concerning the structural characterization of either the entire ␤-amyloid peptide, or of some of its synthetic fragments, in relation to their aggregation properties (for review, see Ref. 2). Many studies have demonstrated that the A␤ peptide, as well as analogs, become toxic for cultured neuronal cells only when they form fibrillar aggregates (8,9). This was observed for the 25-35 and 34 -42 amyloid peptides, which are highly toxic for cells and are both able to form amyloid fibrils. Recent results from Lansbury et al. (6) proposed a model of the formation and structure of fibrils, where the amyloid 34 -42 peptide plays a critical role in the stabilization and growth of these fibrils. However, these structural data could not provide with a mechanism for fibril formation or for the damages induced by the A␤ peptide to cell membranes and for its ability to kill neurons. Recent studies suggest that the amyloid peptide might be toxic for cultured neurons via an oxidative mechanism independent of a receptor interaction, suggesting a direct interaction of the peptide with the cell membrane (14). By analogy with other peptides, active on cell membranes, A␤ has an ordered amphiphilic nature, which could be in part responsible for its toxicity. Another amyloidogenic peptide, calcitonin, appears as an amphiphilic ␣-helix in solution. The cluster of positive charges localized on one side of the helix suggests that the hydrophilic part of the molecule remains at the cell surface while the hydrophobic part of the peptide can penetrate the plasma membrane (32). Similarly, the cytotoxic peptides, mastoparan and mellitin, are thought to insert into the membrane where they perturb several membrane-associated enzymes and signaling pathways (33,34). Fusion of phospholipid membranes is thought to involved three steps: vesicle aggregation, membrane destabilization, and merging of membrane. Using computer modeling, we localized in the C-terminal domain of A␤, a peptide with physico-chemical properties close to those of putative fusion peptide of several viral proteins (20). The obliquity and the high hydrophobicity of the 29 -40 and 29 -42 amino acids of A␤ are consistent with such potential fusogenic properties. Our in vitro data clearly indicate that these peptides are able to induce fusion of unilamellar lipid vesicles. The results from both the lipid-mixing and the core-mixing experiments (Figs. 4,8,and 9), and the monitoring of vesicle size increase ( Fig. 6 and 7) shows that the A␤-(29 -42) peptide is the more potent fusogenic peptide and that its fusogenic properties are closest to those of the SIV fusion peptide. By analogy with this peptide, the A␤-(29 -42) peptide may be a better fusogenic peptide, because it can insert deeper into the lipid bilayer than the A␤- (29 -40) peptide. Considering that in our lipid-mixing assay, the ratio of the fluorescence-labeled/unlabeled vesicles is 1:4, the above decrease in the surface density of the pyrene-labeled phospholipids is consistent with an amyloid peptide-induced fusion between the labeled and unlabeled SUVs, followed by lateral diffusion of the fluorescent probes in the plane of the newly enlarged membranes. Simple aggregation of the vesicles would not result in such a change of the pyrene excimer/monomer ratio (35). Moreover, fused vesicles could be resolved by gel filtration, and large and unspecific aggregation of the lipid vesicles was avoided during the time course of our experiments. When the fluorescence-labeled vesicles were mixed with the peptides in the absence of unlabeled vesicles, there was no change in the pyrene excimer/monomer ratio. This indicates that the effect observed above was not due to a simple exchange of pyrene-phospholipid between labeled vesicles and was not due to a change in the shape or in the radius of curvature of the vesicles. In the same way, mixing of labeled and unlabeled vesicles in the absence of peptide did not result in a decrease of the pyrene excimer/monomer ratio. Thus, the vesicles used were stable for at least 1 week at 4°C, indicating that an apparent fusion resulting from vesicle instability can be discarded (30,36). These results were further confirmed by coremixing experiments (Fig. 9). Within the same concentration range used for the lipid-mixing assay, the four amyloid peptides tested were able to induce mixing of core contents of liposomes. In agreement with the data from the lipid-mixing assay, the A␤-(29 -42) was the most fusogenic peptide and the elongation of this peptide at its N-terminal end did not abolish the fusogenic properties of the C-terminal domain of the amyloid peptide.
This fusogenic activity is related both to the peptide concentration and to the ␣-helical content of the peptides (Figs. 4 and 5). Formation of an ␣-helical structure may promote easier association with the lipid bilayer and penetration of the peptide. Such observations have been described for other fusion peptides such as the N-terminal fusion peptide of the Sendai virus (27). These results are supported by experiments carried out with mutant A␤-(29 -42) peptides with the same hydrophobicity but a different orientation at a lipid/water interface and with the A␤-(13-28) peptide, which is not fusogenic. These data stress the crucial effect of the obliquity and hydrophobicity of the peptide to induce fusion and suggest that the oblique membrane insertion of the A␤ peptides and their fusogenic properties might potentiate toxicity. Elongation of the tilted peptides at their N-terminal end do not significantly modify their fusogenic properties. The A␤-(22-42) and the A␤-(12-42) peptides still induce the fusion of LUVs and SUVs as determined by lipid-mixing and core-mixing assays (Figs. 8 -10) and by visible absorbance measurements (not shown). Since the A␤-(13-28) peptide is not able to cause fusion, these data strongly suggest that the C-terminal end of the A␤ peptide is indeed responsible for the interaction and the destabilization of the lipid bilayer and for further fusion. Moreover, the tilted C-terminal peptide retains its fusogenic properties within the longer peptides, thus providing valuable clues toward the neurotoxic effect of the entire peptide. The decreased fusogenic capacity of the A␤-(12-42) and A␤-(22-42) peptides compared with that of the A␤- (29 -40) and the A␤-(29 -42) peptides, might be due to sterical hindrance created by the folding of the entire peptide. It has been reported that a change in the conformation of a protein could result in the exposure of the fusion peptide. At the pH of fusion, an essentially irreversible conformational change of influenza virus hemaglutinin results in the exposure of the N-terminal fusion peptide (37,38). Experiments are currently performed to determine to what extent the structure of the entire A␤ peptide could be modified by the incubation conditions and to correlate those parameters with the fusogenic activity of the C-terminal domain of the A␤ peptide.
We further investigated the ability of the tilted amyloid peptides to modify the permeability of the SUVs by measuring the leakage of the encapsulated calcein from the vesicles, upon the addition of the peptides. The rate of increase of the calcein fluorescence intensity was taken as an index of the changes in vesicle permeability. There was release of calcein upon addition of the A␤- (29 -40), A␤-(29 -42), A␤- , and A␤-  peptides. Thus, the amyloid peptides induce vesicle fusion with concomitant alteration of the bilayer permeability. On the contrary, the A␤-(12-28) and the A␤-(29 -42) mutant peptides do not cause significant calcein release, suggesting an absence of efficient bilayer perturbation mediated by these peptides. Our results strongly suggest that the C-terminal end of the ␤-amyloid peptide is able to induce liposomal fusion through its deep insertion at a tilted angle into the lipid phase and that fusion. Data from fusion experiments, together with membrane perturbation of the liposomes, inducing calcein release out of the vesicles support the conclusion that fusion of liposomes induced by the amyloid peptides is accompanied by membrane leakage. Moreover, it has been proposed that membrane fusion, which takes place after insertion of the peptide into the membrane, is followed and amplified by the self-aggregation of peptide within the lipid phase, suggesting that peptide oligomerization might play an important role in the fusion induced by polypeptides (39,40). In separate experiments, we have shown that the A␤-(29 -42) peptide aggregates more rapidly than the A␤- (29 -40) peptide in an aqueous buffer and forms amyloid-like fibrils. 2 This is in good agreement with the enhanced fusogenic properties of the A␤-(29 -42) peptide. We therefore propose that the C-terminal domain of the amyloid peptide is able to interact directly with a lipid bilayer, due to its obliquity and hydrophobicity, and to insert into the membrane. This interaction mechanism might account for the perturbation of cellular membranes by the amyloid peptide and for its toxicity toward a wide range of cells. In its soluble form, the amyloid peptide is toxic only at high concentrations, while the extensively aggregated peptide is not toxic. The A␤ peptide, either in its soluble form or when present in diffuse amyloid plaques, can reach high concentrations and be in the appropriate conformation to interact with a lipid membrane through its C-terminal domain. This might cause membrane perturbation and increase the susceptibility of the cell to further damages. According to this hypothesis, the amyloid peptide might be lethal to cells through its interaction with the neuronal plasma membrane, which in turn might activate one or more enzymatic or second messenger systems, and thus lead to cell death. In conclusion, the ␤-amyloid peptide is able to promote the fusion of phospholipids, causing modification of both the lipid structure and the bilayer permeability. Bilayer fusion, and thus insertion of the peptide into membrane might reflect the processes occurring during the interaction of amyloid peptide with neuronal cell membranes and thus it gives an insight into the molecular basis of the cytotoxicity of this potentially toxic peptide.